Post on 29-Jul-2020
1
Chapter 2
Wavefield representation, propagation and
imaging using localized waves: Beamlet, Curvelet
and Dreamlet
Ru-Shan Wu and Jinghuai Gao
This Chapter is a review on the application of wavelet transform to wave propagation and
imaging. The Authors review the development of phase-space localized propagators in
heterogeneous media and the application to geophysical, especially seismic imaging. In
the first part, phase-space localization, mainly along the line of time-frequency
localization is reviewed. Then phase-space localization using generalized wavelet
transform applied to wave field and one-way propagator decompositions are reviewed
and analyzed. Physically the phase-space localized propagators are beamlet or
wavepacket propagators which are propagator matrices for short-range iterative
propagation. When asymptotic solutions are applied to the beamlet for long-range
propagation, beamlets evolve into global beams. Various asymptotic beam propagation
methods have been developed in the past, such as the Gaussian beam, complex ray,
Gaussian packet, coherent state, and more recently the curvelet and dreamlet (tensor
product of drumbeat and beamlet) methods. Local perturbation method for propagation in
strongly heterogeneous media is also briefly described. Curvelet transform and its
application to propagation and imaging is reviewed in comparison with the beamlet
approach. Finally physical wavelet is introduced and dreamlet as a type of physical
wavelet on the observation surface is discussed with the recent applications to seismic
data decomposition, compression, extrapolation and imaging.
Based on the review and analysis, some conclusions are reached as follows. For
wavefield decomposition, beamlet, dreamlet and curvelet transforms have elementary
functions of directional wavelets. Beamlet and dreamlet belong to a type of physical
wavelet, representing an elementary wave (satisfying wave equation) in various
wavefield decomposition schemes using localized building elements (wavelet atoms),
such as coherent state, Gabor atom, Gabor-Daubechies frame vector, local trigonometric
basis function. Curvelet transform is a specifically defined mathematical transform,
characterized by the parabolic scaling. The parabolic scaling law, width length2
or its
generalization width wavelength2 is similar to the beam-aperture requirement for
asymptotic beam solution: the beamwidth must be smaller than the scale of heterogeneity
2
and much greater than the wavelength, i.e. a > β >> λ where is the wavelength, is
the beamwidth, and a is scale of heterogeneity. Optimal beamwidth is reached by
balancing the beam geometric spreading which is controlled by the ratio / , and the
beam-front distortion which depends on a / . Using optimal beamwidth,
beamlet/dreamlet or curvelet propagator will be sparse in smooth media for short range
propagation. For strong and rough heterogeneities, beamlet/dreamlet or curvelet
scattering will occur and asymptotic propagator may not work well. In this case, the local
perturbation method can be applied, in which the propagator is decomposed into a
background propagator and a perturbation operator for each forward marching step.
Numerical examples demonstrated the validity of the approach.
Key words: Beamlet, Curvelet, Deamlet, Wavelet transformation and decomposition,
Localized wave propagators, Gaussian beam method, Wavepacket method
2.1. Introduction
The development of the theory and methods in wavefield decomposition,
propagation, and imaging plays a fundamental and crucial role to seismic exploration, or
more general, to geophysical, including, acoustic, electromagnetic, and seismic,
exploration. In this Chapter, we concentrated on one-way propagation and its application
to imaging. In the past thirty years, various techniques and methods, including high-
frequency asymptotic method (ray method, Maslov method), the wave-equation based
phase-shift method for laterally homogeneous media, phase-screen method (or Split-step
Fourier method) for weak heterogeneous media, Hybrid Dual-domain method (Fourier-
finite-difference method and generalized screen method) for strong heterogeneous media,
have been developed for the applications in different environments. For reviews of recent
progress, see Wu and Maupin (2007). In these methods, the wavefield is expanded by
sets of basic functions such as spatial Fourier harmonics, modes, and spatial Green’s
functions. The common factor in these basic functions is their global nature in either the
space domain or the wavenumber domain. Basic functions in the wavenumber domain
are plane waves, which have the best localization in direction but no spatial localization.
Basic functions in the space domain are point sources (delta functions), which have the
best localization in space but have no direction localization. In highly heterogeneous
media, the global nature of these two kinds of basic functions creates some difficulties in
dealing with strong and fast-varying heterogeneities. To overcome this fundamental
limitation caused by the global nature of these propagators, efforts have been made to
investigate and develop wavefield decomposition and extrapolation methods with
localization in both space and direction.
The newly developed fast wavelet transform (WT), with its extension and
generalization, is considered to be a revolutionary breakthrough in signal
analysis/processing. Its impact to wave propagation and imaging has only started to be
noticed and could never be overestimated. On the other front, there has been significant
3
progress in one-way wave propagation theory and algorithms, including the recently
developed multi-domain techniques, such as the fast acoustic and elastic generalized
screen propagators and their applications to imaging (Wu et al., 2012). The cross-
breeding of these two new developments has the potential of revolutionizing modeling
and imaging techniques for complex Earth media.
In this review paper, we will use the terms “wavelet”, “wavelet transform”, and
“wavelet decomposition” in a very general sense. “Wavelet transform” means a transform
using a phase-space localized, multi-scaled (single-scale is a special case) elementary
function. In the case of classical wavelet transform using a basis generated by translation
and dilation of a mother wavelet, we will refer it as the “classic” wavelet transform.
The distinctive and fundamental feature of wavelet transform (generalized) is so
called “the multi-scale phase-space localization”. In the context of wave propagation, the
phase-space localization means the localization in both the space (space-domain) and
propagation direction (wavenumber-domain). The phase-space localization gives the
possibility and flexibility in treating the fast-varying lateral heterogeneities in wave
propagation. We show the comparison between the global perturbation and the local
perturbation methods for the SEG-EAGE salt model (Fig. 2.1) to demonstrate the
advantages of localized propagators in Fig. 2.2.
The original velocity model is decomposed into a bi-scale model: A large-scale
reference medium with the scale of window-width defined for spatial localization (Fig.
2.2c), and the small-scale perturbations with respect to the reference medium (Fig. 2.2d).
Compared with the global reference medium (Fig. 2.2a) and the global perturbations (Fig.
2.2b), it is clear that the local perturbations are much weaker and are small-scale in
nature, since the large-scale, strong velocity variations are modeled by the reference
medium. In this example, we see the advantages of the space localization. Equally
important is the direction localization. With the direction localization, we can apply some
angle-related operations to the localized wavefield. We know that, the space-domain
representation of a wavefield has the maximum space localization; however, it does not
have any direction localization. The direction localization of a space-localized wavefield
can have a direction-resolution obeying the Heisenberg uncertainty principle: the location
and the direction of a wave field cannot be simultaneously given exactly, and the product
of location uncertainty and direction uncertainty must be equal or greater than a
minimum value: the Heisenberg cell. In contrast, under the h-f asymptotic theory (ray
theory), both the space and direction can be specified with infinite accuracy. However,
this assertion does not reflect the physical reality of wave phenomena and will give
erroneous predictions for wave propagation in complex media.
Fig. 2.1 2D SEG/EAGE velocity model used in this study.
4
Fig. 2.2 Comparison between global perturbation and local perturbation schemes: (a) Global reference velocity
for global perturbation methods; (b) Velocity perturbation with respect to the global reference model; (c) Local
reference velocity for the beamlet method; (d) Local velocity perturbations for beamlet method.
2.2. Phase-Space Localization and Wavelet Transform
Historically, phase-space localization has been developing along two lines: the time-
frequency (or space-frequency) (x-ξ) localization and the time–scale (x-a) (or space-
scale) localization. Here “x” is the time or spatial axis, and “ξ” is the time or spatial
frequency, and “a” is the scale. In the context of wave propagation in this paper, we will
refer x as the variable (may be multidimensional) in space domain, and ξ as the spatial
frequency or wavenumber with the understanding that (x-ξ) means both (space-
frequency) and (time-frequency). Modern revolution in signal analysis has been ignited
by the introduction of “wavelet” and its fast transform algorithms. Wavelets were
introduced in the form of time-scale (x-a) localization, first by a geophysicist J. Morlet
(Morlet, 1982a, b), and later rigorously defined by a physicist A. Grossmann and J.
Morlet (Grossmann and Morlet, 1984). However, the time-frequency localization has
been developed much early in physics and engineering. They were popular in signal
processing and physics, however, as pointed by Jaffard, Meyer and Ryan (2001),
neglected by mathematicians for a long while:
“The former (time-frequency analysis) is a result of cross-fertilization between
signal processing and quantum mechanics, and mathematicians took little interest in
these ideas until recently. The latter (time-scale analysis) was pioneered by
mathematicians long before it was adopted in physics and signal processing.”
5
First, let us review the concept and historical development of the time-frequency
localization since it is more related to our work on wavefield decomposition and
propagation.
2.2.1 Time-Frequency Localization
D. Gabor (1946) first introduced the concept of decomposing signals with localized
Fourier transforms for nonstationary signals. Originally, the localization proposed by
Gabor is realized by the Gaussian window and the sampling rate is critical (no
redundancy). His idea is to divide a wave into segments and to use one of these segments
as the analyzing element. Here we follow the exposition of Jaffard-Mayer-Ryan (2001)
for the mathematical definition. Assume the signal is segmented with window function
( ),w x lb l ∈Z, where b is the nominal length of the window. Then the Fourier
transform is applied to the segments with the calculation of the coefficient
( ) ( )i xe w x bl s x , where ( )s x is the signal. This action is the same as taking the
scalar products of the signal s with the “wavelets”
, ( ) ( )ikax
k lw x e w x bl (2.1)
Where a = Δ Therefore, the Gabor decomposition equation (2.1) can be thought as the
early proposal of wavelet decomposition using the time-frequency atoms. However, the
reconstruction of the Gabor expansion was not well studied at that time. In late 1940’s,
after Gabor’s proposal, many scientists, including L. Brillouin and J. von Neumann,
though that the system defined in equation (2.1) could be used as a basis to decompose
any function in L2(R). Two physicists, Balian (1981) and Low (1985) proved
independently that this is not true. Later G. Battle (1988) proved this theory from the
Heisenberg uncertainty principle (see also Daubechies and Janssen, 1993; Chapter 2 of
Feichtinger and Strohmer, 1998). Mathematically the Balian-Low obstruction can be
stated as the following (Mayer, 2001). If the two integrals 22(1 ) ( )x w x dx and
22(1 ) ( )w d are both finite (the window function is sufficiently regular and
well-localized), the functions , ( ),k lw x k, l∈Z, cannot be an orthonormal basis of L2(R).
In other words, if the wavelet w(x) is an orthogonal basis, then the Heisenberg product,
i.e. the time-frequency resolution-cell (uncertainty), must be infinite. In the case of
Gabor’s critical sampling, the frequency spreading is infinite: no frequency localization at
all. This seemingly insurmountable obstacle gave a blow to the enthusiasm of the search
for orthonormal local Fourier basis, and had blocked the way of finding efficient time-
frequency localization method for many years. To overcome the instability of Gabor’s
reconstruction, studies were focused on two approaches: one is to develop orthonormal
bases that can evade the Balian-Low obstruction; the other is to develop frame
representations which are not orthogonal but still efficient.
6
2.2.1.1 Wilson-Malvar Bases and Trigonometric Bases
This approach was initiated by a physicist Wilson (Nobel laureate) (Wilson, 1971)
when working in quantum mechanics. He observed that one does not need basis functions
that distinguish between positive and negative frequencies of the same order when
studying kinetic operator. He proposed to replace the exponential function with the
cosine and sine functions alternatively. Later Malvar (1990), a scientist in signal
processing, discovered a similar approach independently. Significant improvement along
this line has been made (Daubechies et al., 1991; Coifman and Meyer, 1991;
Wickerhauser, 1993, 1994) and the local cosine/sine bases with adaptive window lengths
were constructed. The local trigonometric (cosine/sine) bases are orthonormal bases, and
are also called local Fourier bases (Auscher, 1994). The evasion of Bailian-Low theorem
through the replacement of exponential basis function with trigonometric functions can
be explained as following (Auscher, 1994). The trigonometric functions can form an
orthonormal basis if and only if { , ( )k lw x } is a tight frame with redundancy 2
(Daubechies et al., 1991). “A tight frame with redundancy 2 contains twice as many
vectors needed to form a basis and the linear combinations eliminate this redundancy to
actually yield an orthonormal basis”. However, trigonometric bases do not have the
complete directivity localization as the local Fourier decomposition (WFT or windowed
Fourier frame). The local cosine basis function has always two symmetric spectral lobes:
a positive lobe and mirror-symmetric negative lobe. This can be seen from equation
(2.14) in the next section. Therefore the directivity is not uniquely specified. Although a
local exponential function can be formed by combination of cosine basis function in the
real part and sine basis function in the imaginary part, there is always some spectral
leakage occurs at the window overlapping areas, as shown by Wickerhauser (1994,
Chapter 4.1). That is to say, there is always some false energy in the negative mirror
direction if the real energy has only a positive lobe. The amount of leakage is
proportional to the overlapping area and therefore inversely proportional to the steepness
of the bell side-slope. With steeper side-slope, the leakage will be smaller. On the other
hand, the Heisenberg uncertainty cell-size is proportional to the steepness of the bell
slope. With steeper side-slope, however, the Heisenberg cell-size will be greater, so the
directivity resolution will be poor. When the steepness is infinite, there will be no spectral
leakage, but the spectral spreading is infinite and there is no direction localization!
Therefore, the trigonometric bases are not ideal for directivity representation. There is a
tradeoff between the leakage and resolution. As for direction localization, the “lunch” is
not totally free and you have to pay something. The trick is to trade the important thing
you want with things not valuable to you. Even though not ideal for direction
localization, the trigonometric bases have been successfully applied to wave propagation
as localized propagators for imaging due to the efficiency, which will be reviewed in the
next section.
2.2.1.2. Windowed Fourier Frames
7
Traditional windowed Fourier transform (WFT) uses densely overlapping windows,
which can be any type of smooth function. (e.g. Daubechies, 1992; Kaiser, 1994).
However, the implementation of the WFT is time-consuming and prevented it to be
useful in practical applications. The development of the frame theory provided an
efficient approach for discretizing the WFT. Frame representation is complete but not
orthogonal, and therefore has redundancy in the representation. Daubechies (1990) has
proved that the reconstruction using critical sampling is unstable, and for stable
reconstruction over sampling (x < 2) must hold, where x measures the size of
windowed Fourier atoms (latticed coherent states). The necessary and sufficient
conditions for the stable reconstruction have been derived based on the frame theory
(Daubechies, 1990; 1992). The dual frame vector for reconstruction (synthesis frame
vector), which in general is different from the original frame vector for decomposition
(decomposition frame vector) can be calculated efficiently (Mallat, 1998; Qian and Chen,
1996). The frame vector with a Gaussian window is called the Gabor frame or windowed
Fourier frame. It is also called as Weyl-Heisenberg coherent state frame (Daubechies,
1990) because of its relation with the Weyl-Heisenberg group in quantum field theory
(Klauder and Skagerstam,1985; Foster and Huang, 1991) or Gabor-Daubechies (G-D)
frame (Wu and Chen, 2001, 2002a, Chen et al., 2006).
Ville’s work interprets the time-frequency localization in terms of pseudo-
differential operators. This started from Weyl’s formulism and the work in quantum
mechanics of the physicist E. Wigner in the 1930’s. The technology transfer to signal
processing begins with Ville’s work in 1950’s.
2.2.2 Time-Scale Localization
Fourier transform was and still is considered to be a revolution in the history of
signal analysis. Any (or almost “any”) signal (function) can be decomposed into various
harmonic (wave-like) components. However, Fourier analysis, namely the frequency
analysis, is not convenient for scale analysis, especially for multi-scale analysis. The
Fourier decomposition is not efficient for signals with discontinuities or abrupt changes.
There was a need for another revolution for efficient multi-scale analysis. The first multi-
scale analysis with an orthogonal basis was introduced by Haar in 1909, more than 100
years after Fourier invented his analysis (1807). From Fourier to Haar, total new concepts
were introduced to mathematics, and total new weapons of frequency and scale analysis
were brought to the scientists and engineers for attacking different application problems.
However, Haar’s basis is the most discontinuous basis, composed of positive and
negative box functions with different widths (scales), and is not efficient, even not
appropriate to decompose continuous functions. In comparison, the Fourier basis is
composed of most continuous functions, the sine and cosine functions. The revolution of
multi-scale analysis can only be said is really succeeded until Daubechies orthonormal
bases were discovered. Daubechies basis function is regular and has continuous
derivatives up to certain degree. With the new-generation basis functions, the multi-scale
analysis becomes an efficient and powerful tool for scientists and engineers and many
tough problems which are untreatable before can be attacked now.
8
Jaffard et al. (2001) have an excellent review on the historical development along the
line of time-scale localization. The authors (JMR) divided the development into three
periods. In the first period, wavelets had been studied individually by some
mathematicians, from Lévy, Littlewood and Paley, Franklin, Lusin, Calderόn, Coifman
and Weis, to Strőmberg. In this period, those mathematician’s works are unknown to the
physics and signal-analysis communities and did not form a unified approach. The
second period started with a more general definition and unified approach of wavelet
(Grossman and Morlet, 1984) and a broad application of the theory to physics and signal
processing. In fact the term “wavelet” was first introduced by Morlet et al. (1982a,b). As
JMR pointed out, although the basic theory of Grossman and Morlet (1984) was a
rediscovery of Calderόn’s identity (1960), but the synthesis and unification of the scale
analysis by physicist Grossman and geophysicist Morlet, had a powerful push for the
development of the mathematics of wavelets. JMR called this as the first synthesis. The
third period, they called the second synthesis, is marked by the unification of the works
by not only the mathematicians but also the works by scientists/engineers working in
signal/image processing, digital communication, and physics. This is the period of
modern wavelet theory, the most flourishing period of the theory. Daubechies discovery
was inspired by the relation between wavelets and the quadrature mirror filtering.
In the conclusion of their historical review, JMR has an interesting comment, which
we cited here:
“Today the boundaries between mathematics and signal and image processing have
faded, and mathematics has benefited from the rediscovery of wavelets by experts
from other disciplines. The detour through signal and image processing was the most
direct path leading from the Haar basis to Daubechies’s wavelets.”
2.2.3 Extension and Generalization of Time-Frequency, Time-Scale
Localizations
In spite of the success of discrete wavelet transform based on time-scale localization
and its outstanding features in signal analysis, it has some shortcomings in dealing with
long oscillatory signals, such as the textured images, fingerprints, and wave phenomena.
Meyer (1998) has an excellent comment on the strong and weak points of time-scale
approaches:
“Wavelets, however, are ill suited to represent oscillatory patterns. Rapid variations
of intensity can only be described by the small-scale wavelet coefficients. Long
oscillatory patterns thus require many such fine scale coefficients. Unfortunately
those small scale coefficients carry very little energy, and are often quantized to
zero, even at low compression rates. “
Here “wavelet” is referred to the classical wavelet in our terminology. On the other hand,
time-frequency localization has some disadvantages in representing the multi-scale
structures with sharp, irregular boundaries, such as those encountered in image
processing. It was the consensus in the community of sciences and engineering that
9
neither of the two approaches can be optimally decomposes all the signals/images in
different fields. Realizing the limitations and restrictions of two approaches, investigators
tried to generalize the two categories of phase-space localizations.
Wavelets as representations of groups based on the group theory:
From the viewpoint of group theory, the classic wavelet transform based on time-
scale localization is related to the affine group, which is invariant of translation and
dilation. The windowed Fourier transform and windowed Fourier frame based on time-
frequency localization are related to the Weyl-Heisenberg group, which is in variant with
time and frequency shifts. Wavelets can be considered as the canonical representation
(square-integrable) of the corresponding groups.
The extension can be done by adding frequency shift in the affine group or adding
dilation in the Weyl-Heisenberg group, resulting in an affine Weyl-Heisenberg group
(Torrésani, 1991). In this way, the flexibility of selecting best bases is increased,
resulting in more efficient representations for complex image structures. In wavefield
decomposition, this generalization provides the possibility of more efficient
representation of complicated wavefield and wave propagator (See section 2.3 for
details).
Another direction of extension is proposed by Antoine et al. (1996) by adding the
rotation action to the affine group. Taking all the actions together, these transformations
are represented by a unitary operator ( , , )x a in the space L2(R
2,d
2x) of signals. The
three operations of translation, rotation and dilation generate the similitude group SIM(2)
of R2, i.e. the two-dimensional Euclidean group with dilations.
The scale-angle pair may be interpreted as spatial frequency variables in polar
coordinates. Directional wavelets were introduced as wavelets whose Fourier transform is
(essentially) supported in a convex cone in spatial frequency space, with apex at the
origin. In physical terminology, the 2-D wavelets are the coherent states associated to the
representation of SIM(2) (Torrésani, 1991). Fig. 2.3 shows the tiling of the spatial
frequency plane with the effective support of the wavelet.
Definition of the unitary operator:
1 1
, ,( ( , , ) )( ) ( ) ( ( ))aa s x s x a s a r x
bb b (2.2)
Or in frequency (k) domain
, ,
ˆ ˆ ˆ ˆ( ( , , ) )( ) ( ) ( ( ))i
aa s s ae s ar
b k
bb k k k (2.3)
And the directional wavelet:
2
22
2ˆ(2 ) ( )
dc
kk
k
(2.4)
The selection of the spatial frequency vector k: The best angular selectivity is
obtained by taking the vector perpendicular to the largest axis of the modulus, thus
10
0(0, )kk . This selection is same as the definition of curvelet and wave atom discussed
in section 2.4.
Fig. 2.3 Tiling the spatial frequency plane with the effective support of the wavelet (From Torrésani, 1991)
The above extensions are defined for continuous wavelet transform (CWT), the
numerically implementation and fast algorithm have been searched from two directions.
One is the projections method with a tree structure from a dictionary or library (optimum
wavelet packets) (Coifman and Wickerhauser’s best basis algorithm). The other is the
matching pursuit (or other pursuit schemes) to find the “best” approximate representation
(Mallat, 1989; Mallat and Zhang, 1998). The third method is the Wigner-Ville transform
(see Chapter 5 of JMR) (its relation to the Cohen class, see Cohen, 1995).
From the above summary on the extensions of the two kinds of phase-space
localization, we see that the generalizations from different approaches have not
converged to a common terminology. This can be seen from the increasingly large
number of terms introduced by different investigators under different schemes for
different applications, such as ridgelets (Candės, 1999; Donoho, 1995), brushlet (Meyer
and Coifman, 1997), vaguelette (Donoho, 1995), curvelet (Candes and Donoho, 1999),
chirplet (Baraniuk and Jones, 1990; Coifman et al., 1997), seislet (Fomal, 2006), wave-
atom (Demanet and Ying, 2006), physical wavelet (acoustic wavelet) (Kaiser, 1994),
eigen-wavelet (Kaiser, 2004)), among others. Things can be even more divergent, since
different researchers can design their own “some-lets” by a lifting algorithm (e.g.
Sweldens and Shröder, 1996) to fit their special signals. Borrowing one sentence from
Jaffard, Meyer and Ryan (2001), “it appears that things are in a state of disorder and
11
confusion”. This appears also natural, since signals are so different in nature, no single
type of bases or algorithms can be optimum to the decomposition of all the signals. In the
following section, we will concentrate on the decomposition of wavefield and the related
propagators (operators), and summarize the recent progress in this field.
2.3 Localized Wave Propagators: From Beam to Beamlet
Despite the confusion and divergence in the terminology relating to localization
elements of phase-space in the field of signal processing, however, the phase-space
localization has a very clear and well-defined physical meaning in wavefield
decomposition and propagation. In order to decompose wavefields efficiently, the
elementary function needs to have the ability of representing oscillating and directional
signals. This is the property of “wave beams”. Beam physics has been studied since the
early time for directional, localized wavefield in seismology, such as Gaussian beam
(Popov, 1982; Červený et al., 1982; Červený, 1983, 1985, 2001; Hill, 1990, 2000;
Albertin et al., 2001, 2002, Nowack et al. 2006) in radar, optics, and other fields of
physics and engineering. The beam properties are also studied under the term of complex
ray, complex source (Kravtsov, 1967; Deschamps, 1971; Keller and Streifer, 1971;
Felsen, 1975, 1976, 1984; Wu, 1985), coherent states (Klauder, 1987; Foster and Huang,
1991; Albertin et al., 2001; Foster et al., 2002). However, the decomposition becomes an
efficient tool for wavefield representation and propagation can only be achieved after
wavelet transform was introduced into physics and geophysics (wave physics).
The term “beamlet” is a general physical term, first introduced by Mosher, Foster
and Wu (1996). Originally it meant to represent any elementary function of decomposing
a wavefield by wavelet transform (in the generalized sense) along the spatial axes.
Therefore it is a space-wavenumber (spatial frequency) localization atom with scaling
capability (dilation). Beamlet representation of a wavefield along a decomposition plane
differs from the traditional space-domain or wavenumber-domain representations at the
dual localizations in both spatial location and propagation direction (phase space
localization): each element of the representation is a beamlet, i.e. a “small beam”.
Beamlet decomposition is also different from the decomposition using Gaussian beams
which are approximate in decomposition and high-frequency asymptotic in propagation.
Different wavelet atoms have been tested for wavefield decomposition and propagation
(Wu and Yang, 1997; Wu and Wang, 1998, 1999; Wang and Wu, 1998, 2000, 2002;
Chen et al., 2001; Chen and Wu, 2002; Foster et al., 2002; Mosher et al., 2002). It
became apparent that only smooth wavelet atoms, either orthonormal bases or frame
vectors, can efficiently represent wavefields and their propagation in heterogeneous
media. Many popular wavelet bases in imaging processing, such as the Daubechies
wavelets, do not work well for wave propagation. The beamlet atoms need to be well
localized in both the space and frequency domains. From the viewpoint of the group
theory as we discussed in the previous section, beamlets can be defined as space-
wavenumber (time-frequency) atoms related to the finite-energy (or square-integrable)
representation of the affine Weyl-Heisenberg group. In the case of CWT (continuous
wavelet transform), the beamlets are space-shift, wavenumber-shit and dilation invariant.
In the case of discrete transform, the shift and dilation invariants are understood in the
12
sense of latticed invariant. In the special case of wavefield representation, its similarity to
the representation of the similitude group can be seen from the relation of the local
wavenumber vector k to the local horizontal wavenumber , where the bar “ ” on top
of a parameter stands for the localized (phase-space localization) parameter. is the
localized frequency along the decomposition axis (in the 2D case) or decomposition
plane (in the 3D case):
( , ) k (2.5)
where ( , )x yk k is the local horizontal wavenumber in the decomposition plane, and
2
2(2 / ) (2.6)
is the local vertical wavenumber, and 2 /c is the wavelength of the localized
wavefield with c as the local wave speed and ω as the circular frequency of the
wavefield.
The beamlets defined in this way can be related to the directional wavelets (defined
as the representation of the similitude group) by the correspondence between the local
rotation angle and the local wavenumber
0sin /2
k
, (2.7)
with 0 /k c .
The decomposition plane is designed to perpendicular to the main (dominant)
propagation direction, which can be determined by some asymptotic methods, such as ray
tracing or Hamilton-Jaccobi equation. In case of short range propagation with data
acquired on the surface of the earth, such as in exploration seismology, the dominant
direction is often preset to the depth (z) direction. Along the main propagation direction,
the beamlet with 0 0(0, )kk is same as the directional wavelet defined by Antoine et al.
(1996). The other beamlets are obtained by wavenumber (ξ) shift from 0k . For the
directional wavelet they are generated by rotation of the 0k wavelet, and for each wavelet
the oscillating component is always perpendicular to the beam-width (phase-front).
Therefore, in the context of wave propagation, beamlets and directional wavelet have the
same capacity of presenting phase-space localized wavefield. This is due to the unique
feature of wavefield: the angle-wavenumber correspondence. However, there are some
differences between these two decomposition schemes which may have different
consequences for wave propagation:
1. Beamlets are decomposed on the data-plane (observation plane) or extrapolated
data-plane, while the directional wavelets are decomposed in the whole image
space. This difference can be seen more clearly from the comparison between
13
dreamlet and curvelet (see Figure 2.20 and 2.21). These two decompositions
may have different noise-resistant properties. The decomposition of directional
wavelets may be less prone to the noise contamination.
2. The sampling in the angle-domain is different. Since
0
0
sin
cos
k
k
(2.8)
Therefore the sampling in the local angle-domain is nonuniform for the beamlet
decomposition.
3. Wave propagation will be formulated differently for these two kinds of
decompositions. Beamlets are very convenient for one-way wave propagation
(wavefield extrapolation), while the propagation of directional wavelets is
formulated by the movement of these wavelets according to the wave-equation
under certain approximation. This will become clear in the discussion of
curvelet propagation in the next section.
2.3.1. Frame Beamlets and Orthonormal Beamlets
Beamlet decomposition can be realized using either orthonormal bases or
overcomplete frame vectors. Windowed Fourier frame (WFF), especially the Gabor
frame which is a Gaussian-shaped WFF, is a widely used wavefield decomposition atom.
Gabor (1946) originally represented signals with sampling intervals satisfying ΔxΔξ = 2π,
where Δx and Δξ are the space (time) and wavenumber (frequency) sampling intervals,
which was later proven to be the most compact representation of this kind (Daubechies,
1990, 1992) and therefore is an orthogonal decomposition. However, Daubechies later
proved through the frame theory that the reconstruction under Gabor’s critical sampling
is unstable and that oversampling ΔxΔξ < 2π must hold for stable reconstruction
(Daubechies, 1990, 1992). Optimally localized in both space (time) and wavenumber
(frequency) domains under the Heisenberg uncertainty principle, the Gaussian window is
the most favorable for windowed Fourier analysis of signals (Mallat, 1998). Generally,
the reconstruction frame vectors (the dual frame vectors) are different from the
decomposition frame vectors, and therefore are no longer Gaussian. However, in the case
of tight frame or nearly tight frame, the dual frame vectors are the same or nearly same as
the decomposition vectors. The windowed Fourier frame with a Gaussian window was
named by Daubechies as the Weyl-Heisenberg coherent state frame. It was also called the
Gabor frame (Feichtinger and Strohmer, 1998) and the Gabor-Daubechies (G-D) frame
(Wu and Chen, 2001, 2002a, Chen et al., 2006).
The elementary functions (vectors) of the G-D frame decomposition are Gaussian
enveloped (Gaussian windowed) harmonic wavefields:
( ) ( ) ( ) mim x i x
mn x ng x g x n e g x x e (2.9)
14
where g(x) is a Gaussian window function, xn nx is the space locus and mm ,
the wavenumber locus of the beamlet. Applying G-D frame decomposition to the
wavefield in a local homogeneous region is like decomposing the wavefield into many
small Gaussian beams (beamlets) at different window locations and with different beam
directions. As shown in equation (2.5), the local wavenumber mm of a
decomposition atom can be associated to a horizontal wavenumber of a beamlet (a
directional wavelet atom), and a propagating beamlet can be expressed as
( )
( ) ( ) m mi x z
mn nb x g x x e
(2.10)
With as the vertical wavenumber defined by equation (2.6). Here z means a small
step of extrapolation. The direction defined by ( , )m m is the nominal direction of a
beam-lobe. The angular spectra of the lobe can be seen more clearly from the
wavenumber domain expression of the beamlet
( )
( , ) ( ) nn mi xi x x i z
mn mb x z d e g e
(2.11)
From the elementary wave expressed by (2.9) we can say that the G-D beamlet is
equivalent to the local Fourier beamlet defined by windowed Fourier transform
(Steinberg and McCoy, 1993; Steinberg, 1993; Steinberg and Birman, 1995) and coherent
state atom (Albertin, 2001a,b). However, when we consider the lattice structure of
representation and the reconstruction atoms, the differences appear. The reconstruction
atom of the G-D frame beamlet is the dual frame vector
( ) ( ) mi x
mn ng x g x x e
(2.12)
where g is the dual window function. The local Fourier atom for reconstruction is the
same as for decomposition but the reconstruction of windowed Fourier transform is very
time-consuming.
Beamlet decomposition can be also done with orthonormal bases. In the literature,
the only type of basis function used for wave propagation is the local trigonometric basis:
Local cosine basis (LCB) or local sine basis (LSB) (Wu and Wang, 1998, Wang and Wu,
2002; Luo and Wu, 2003; Wang et al., 2003, 2005). The basis function of LCB can be
written as
2 1
cos2
nmn n
n n
x xx B x m
L L
(2.13)
15
Where nnn xxL 1 is the nominal length of the window, nB x is a bell function
which is smooth and supported by the compact interval 1[ , ]n nx x for
1 ]n nx x , with , are the left and right overlapping radius, respectively.
Fig. 2.4 shows an example of bell functions and basis function (atom) of LCB, where the
left and right slopes are symmetric, i.e. = . The basis function of LCB in equation (2.13) can be also expressed in the Fourier
domain:
1 2
2n m n mix ix
mn n m n m
n
e B e BL
(2.14)
Similar to equation (2.10), we can associate a propagating LCB beamlet to the basis
function:
( ) ( )1{ }
2
n nm m mi x x i x x i z
mn n
n
b x B x e e eL
(2.15)
where
1
2m
n
mL
(2.16)
is the effective wavenumber. We see that, a LCB beamlet corresponds to a two-lobe
symmetric beam.
16
Fig. 2.4 Bell function and basis function of Local cosine basis (LCB).
2.3.2 Beamlet Spreading, Scattering and Wave Propagation in the
Beamlet Domain
Beamlet propagation observes wave equation. Historically, the propagation of
Gaussian beam is developed under the h-f asymptotic theory (Červený, 1983, 2001) for
smoothly varying media. In the case of complex, multi-scaled heterogeneous media,
perturbation methods can be applied in combination with the asymptotic methods. Beam
scattering has been studied in radar, acoustics, optics, quantum mechanics and other
branches of physics and engineering. For the purpose of wavefield extrapolation and
imaging, the forward-scattering problem, i.e. the one-way propagation problem, is more
relevant and useful. In the following we will concentrate on the forward scattering and
propagation problems.
17
The one-way propagator (wavefield extrapolator) can be expressed as surface
integral with space-varying kernel (Green’s function). In the frequency-space (f-x)
domain, the scalar wave equation (Helmholtz equation) can be written as,
2 2 2 2[ / ( , )] ( , ) 0x z v x z u x z (2.17)
where denotes frequency, ),( zxv is wave velocity at (x, z) and ),( zxu stands for
the frequency domain wave field. Here we consider the transversal coordinate x as the
horizontal coordinate in the 2D case, and as 1 2( , )x x in the 3D case. Therefore, in the 3D
case it is understood that1 2
2 2 2
x x x . If the field is known on a give surface, such as
the measurement surface (data surface), then the field at any point surrounding by the
surface can be calculated by the representation integral (Kirchhoff integral)
( ', ') ( , ; ' ')
( , ) ( , ; ' ') ( ', ')s
u x z g x z x zu x z ds g x z x z u x z
n n
(2.18)
where g is the Green’s function, / n is the normal derivative on the surface, and S is
integration surface. If the surface S is a flat surface, the integral can be reduced to a
Rayleigh integral
( , ; ' ')
( , ) 2 ( ', ')s
g x z x zu x z ds u x z
n
(2.19)
The representation integrals are formal solutions of the corresponding wave
equation. Since the integrals (2.18) and (2.19) represent the propagation effects ruled by
the wave equation, these integrals are also called the propagation integrals. These
integrals can be considered as a linear operator P
( , ) ' ( , ; ', ') ( ', ')Pu x z ds K x z x z u x z (2.20)
where ( , ; ', ')K x z x z is kernel and
( , ; ' ')
( , ; ', ') 2g x z x z
K x z x zn
(2.21)
in the case of propagation integral (2.19).
In principle, the propagation integrals can be applied to arbitrarily heterogeneous
media, if we use the exact Green’s function. However, in many cases the exact Green’s
function is not known or too complicated to compute. In many practical applications,
especially in wave-theory based imaging methods, we prefer to neglect the backscattering
with respect to the main propagation direction and the propagation integrals become
extrapolation integrals, which are one-way propagators.
18
If we consider the action of P on a beamlet (wavelet atom of a wavefield
decomposition), it becomes a scattering problem of beamlet. Gaussian beam scattering
has been studied since its introduction (Choudhary and Felsen, 1974; Felsen, 1976;
Červený, 1983; …). Scattering of elementary wavelet has also been studied by Young
(1993). For the extrapolation integrals, mainly the forward scattering problem is
interested. After propagation, the beamlet becomes spread and distorted due to geometric
spreading and scattering, and is no longer a beamlet. If the deformed beamlet is re-
decomposed into beamlets, we can get a propagator matrix in the beamlet domain:
, , ,jl mn jl mn jl mnP b Pb b dsKb (2.22)
where <b,a> stands for the projection of function a onto the wavelet vector b. This matrix
represents completely the linear operator (Candès and Demanet, 2003). In matrix form,
the operator decomposition can be written as (Wu and Yang, 1997; Wu and Wang, 1998)
TP BPB (2.23)
where P is the space domain propagation matrix, and P is the one in beamlet domain; B
is the beamlet vector and T
B is its transpose. Fig. 2.5 illustrates the concept of propagator
decomposition and wave propagation in the beamlet domain.
For beamlet scattering problem, two approaches widely used in wave propagation
theory can be adopted to this purpose: one is the asymptotic approximation for smooth
media; the other is the perturbation method for rough heterogeneities. Traditionally the
Gaussian beam method adopted almost exclusively the asymptotic approach (see
Červený, 1983, 2001); while the beamlet imaging uses the combination of asymptotic
method for the background propagation and perturbation approximation for rough
heterogeneities (Wu et al., 2000; Wu and Chen, 2001, 2002a; Chen et al., 2006; Wang
and Wu, 2002; Luo and Wu, 2003; Wang et al., 2003, 2005).
In the case of smooth media, the local homogeneous approximation or asymptotic
approximation can be applied to the propagator decomposition. Under these
approximations, the Green’s function is known or can be approximated by asymptotical
solutions. Then the propagator matrix in beamlet domain can be easily calculated.
Different wavelet atoms, such as Daubechies wavelet D4, Coifman wavelet C5, best
bases of wavelet packet, and local cosine basis have been tested for the propagator P
decomposition (Wu and Yang, 1997; Wu et al., 1997; Wu and Wang, 1998; Wang and
Wu, 1998a, b). Figs. 2.5 to 2.9 show the decomposition of propagation integral under
local homogeneous approximation into different beamlet domains and the comparison of
sparseness of the beamlet propagators. First, we see obviously that the propagator
matrices in beamlet domain are quite sparse compared to the dense space-domain ones. It
is also seen that the smooth wavelets (with well-localized wavenumber spectra) can
deliver better behaved propagator matrices than the popular wavelets in imaging
compression such as the orthonormal Daubechies wavelets. Especially the local cosine
basis yields very sparse and well-organized propagator matrices, which are important for
efficient implementation. Later in our research, we mostly adopted the smooth wavelets,
19
such as G-D (Gabor-Daubechies) frame and LCB beamlets for wave propagation and
imaging.
Fig. 2.5 Beamlet decomposition and propagation of wavefield.
Fig. 2.6 Matrix representations of Kirchhoff propagation operators in space domain: dx = dz = 25m, v = 2000m/s, N = 128. Left: 5.9Hz, Right: 25Hz. Only real parts of the complex operators are plotted.
20
Fig. 2.7 Beamlet propagators (propagation operators in beamlet domain) (Daub4: Daubechies-4 wavelet): The left panel is for DWT (discrete wavelet transform), and the right panel is for best basis wavelet-packets. The top
panel is for 5.9Hz, and the bottom panel is for 25Hz. Only real parts of the complex operators are plotted.
21
Fig. 2.8 Beamlet propagators (propagation operators in beamlet domain) (Coif5: Coifman-5 wavelet): The left panel is for DWT (discrete wavelet transform), and the right panel is for best basis wavelet-packets. The top
panel is for 5.9Hz, and the bottom panel is for 25Hz. Only real parts of the complex operators are plotted.
22
Fig. 2.9 Propagator matrices of the LCB background propagator in the beamlet domain (Top panel: f = 5.9 Hz;
Middle panel: f = 25.0 Hz) (Left panel: Real part; Right panel: imaginary part) with operator aperture of Nx=256.
2.3.3 Beam Propagation in Smooth media with High-Frequency
Asymptotic Solutions
Equation (2.22) formally defines the beamlet propagator matrix. There is no essential
difficulty in calculating propagation matrices for smooth media under the local
homogeneous approximation or h-f asymptotic approximation. However, the calculation
of the beamlet propagator in generally heterogeneous media is quite complicated. Various
approximations have been obtained for different applications. Historically, there are
basically two approaches. One is the beam propagation methods which calculate the
evolution of the elementary wave (a beamlet) globally using asymptotic solution of the
wave equation in smooth media, and synthesize the wavefield at the end point by
reconstruction (superposition of all the arriving beams). In this approach, the propagator
23
matrices do not enter into the computation and exist only in the definition. Each beamlet,
the elementary wave, evolves into a global beam in the propagating space and arrives at
the receiving point, contributing to the final summation. The other is the perturbation
approach for wave propagation in the beamlet domain. The propagation of a wavefield is
a step-by-step beamlet propagation and coupling using propagator matrices. At each
step, the laterally varying velocity profile is decomposed into a background velocity
profile and local perturbations. Each beamlet will be spread (in the background media)
and scattered (by local perturbations) into other beamlets. No global beam solution is
used in this approach. The details of this perturbation approach will be summarized in the
next subsection.
2.3.3.1 Conditions for the Application of Asymptotic Analysis
Before going reviewing the global beam approach, let us discuss the conditions of
applicability for asymptotic solutions in inhomogeneous media.
The one-way evolution of a beamlet can be formally written as
xbexa mn
ziA
mnn
(2.24)
where An is the square-root operator (for the case of scalar wave).
),(/ 222 zxvA xn (2.25)
Note that amn is no longer a beamlet due to diffraction and scattering.
Redecomposition of amn into beamlets, leads to the formulation of propagator matrix, as
seen in equation (2.22). For general heterogeneous media, the solution of equation (2.24)
can be quite involved and sometimes only numerical methods are applicable. However,
when the media is smooth in comparison to the wavelength, h-f asymptotic
approximations may be applied to the solution. In equations (2.22) or (2.24), if we
transform the propagator into wavenumber (ξ) domain, the application of propagator
operator to the beamlet can be written as
, , ( , ) ,niA z i x
mn mn mna x z e b x z d P x z e b z (2.26)
Standard h-f asymptotic analysis (Garding, 1987; Červený, 1983, 2001; Candès and
Demanet, 2005; Demanet, 2006) assumes that ( , ) i xP x z e can be approximated
( , , )( , ) ( , , )i x i x zP x z e e x z
(2.27)
where ( , , )x z is the phase term and ( , , )x z , the amplitude term. The phase Ф is
homogeneous of degree one in ξ; the amplitude 0 1 ... with
m
24
homogeneous of order –m in ξ. Physically the above conditions mean that the amplitude
term is a much slowly varying function in space, and the phase term is dominated by the
linear dependence in frequency (wavenumber). With this asymptotic approximation, the
one-way beamlet propagator can be represented by a Fourier integral operator (FIO) If
the phase term in (2.27) can be expressed exactly the linear function of frequency, i.e.
( , ) ( , , )i x i xP x z e e x z (2.28)
and ( , , )x z obeys some constraints
,( ( , ) (1 )m
x x C
(2.29)
for multi-indices α and β. Then the propagator belongs to a type of pseudodifferential
operator (PsDO) and ( , )x is its symbol of order m (type (1, 0) (Grossman, 2005;
Demanet, 2006). The space domain expression of the operator is ( , )x D with D i .
The condition in equation (2.29) requires the fast decay in high frequency of the operator
spectrum, implying smooth variation of the wavefield along x. Here z in equation (2.26)
stands for the coordinate along the propagation direction. For a long distance propagation
using asymptotic methods, it can be replaced with the curved-linear coordinates along the
ray which obeys the eikonal equation for the phase. The amplitude can be determined by
the transport equation.
The mathematical requirements for asymptotic analysis can be translate into wave
propagation regimes in terms of beam and medium parameters. For beam evolution in
inhomogeneous media, there three basic parameters that determine the propagation
regimes: wavelength λ, beam width β, and the scale of the heterogeneity a. First, a >>
and >> must hold for the validity of asymptotic approach. For short range
propagation, the beam asymptotic solution requires also a , namely the beamwidth
must be smaller than the scale of heterogeneities (Steinberg, 1993). This is the geometric
optics (GO) regime of wave propagation (Flatté et al., 1979; Wu and Aki, 1988). If
a , the wave front will be distorted and it is in the diffraction regime (ibid). For long
range beam propagation, there are two more parameters enter into the game: the range R
and the total perturbation strength . The requirement of a >> is replaced by
/ 2Fa r R where Fr is the Fresnel radius along the beam path. This is
understandable, since diffraction regime will be entered if the scale of heterogeneity is
smaller than the Fresnel radius. The other requirement that the total r.m.s. perturbation
strength must not exceed certain value is for the case long range propagation in randomly
heterogeneous media (Flatté et al., 1979, Ch. 8; Aki and Richards, 1980, Ch. 13; Wu and
Aki, 1988), which we will not discuss here.
Therefore, for strongly and roughly heterogeneous media, the h-f asymptotic
approximation may be not applicable and the beamlet propagator (or more general the
linear operators of wave propagation in equations (2.18) and (2.20) cannot be represented
by Fourier integral operators and pseudodifferential operators.
25
There are mainly two approaches related to beam propagation: the Gaussian beam
method and the coherent state method. In fact they are close relatives, but with different
historical origins and different approximations to the asymptotic solution.
2.3.3.2 Gaussian Beam Method
Gaussian beam can be considered as evolved from a parabolic approximation of the
Gabor beamlet (or Gabor-Daubechies beamlet) in equation (2.30).
2
00
( )2
( ) ( )
mmi x i k z
kg
mn nb x g x x e
(2.30)
where Δz implies that the wave front is defined in the vicinity of the reference point.
Gaussian beam is a wave-beam with a Gaussian envelope and a parabolic wave front. Its
evolution in smoothly varying media has been extensively studied and documented
(Červený et al., 1982; Popov, 1982; see Červený 2001, and Popov 2002 for detailed
expositions; Červený et al., 2007 for a review of recent progress). The beam propagates
along a ray-path. The spatial trajectories and the travel times are determined by the
eikonal equation (ray-tracing) and change of the beam shape and amplitude is governed
by the dynamic ray-tracing equation which calculates the complex-valued second
derivatives of the travel time along the central ray. In this way, the traveltime and
amplitude in the paraxial region about the ray can be determined. The evolution of a
Gaussian beam depends on two initial parameters, the ray curvature and the beam width.
The Gaussian beam summation provides regular wavefield everywhere even in media
where caustics and shadow zones may be present. The solution for a single beam can be
constructed in f-x domain (e.g., Červený et al., 1982) as well as in t-x domain using a
Hilbert transform (e.g., Hill, 1990, 2000). In seismic imaging, the recorded wavefield is
first decomposed using a set of overlapping Gaussian windows. For each windowed data
section, the local slant stack is performed to form individual beams (e.g., Raz, 1987;
Hale, 1994). Such decomposition is closely related to the decomposition using windowed
Fourier transform and efficient reconstructions can be obtained based on the “frame”
theory (e.g., Daubechies, 1992; Kaiser, 1994) in wavelet analysis. The Gabor frame-
based overcomplete representation provides redundant yet stable reconstruction of the
wavefield (e.g., Einziger, 1986; Nowack et al. 2006). The accuracy of the beam solution
depends on the beam width. High-frequency asymptotic analysis can be also applied to
the non-approximated Gabor beamlet, and the resulted propagating beam is the
windowed Fourier beam (Steinberg, 1993; Steinberg and Birman, 1995), the coherent
state beam (Klauder, 1987; Foster and Huang, 1991; Thomson, 2001; Albertin et al.,
2001; Foster et al., 2002) or the Gabor-frame beam (Gao et al., 2006). Červený et al. also
called the parabolic-approximated Gaussian beam as paraxial Gaussian beam, and the
non-approximated beam as strict Gaussian beam (Červený, 2007).
2.3.3.3 Asymptotic Coherent-State Solutions
26
The original definition of CST (coherent state transform) is a windowed Fourier
transform with Gaussian windows (Klauder and Skagerstam, 1985; Klauder, 1987; Foster
and Huang, 1991). Each Gabor atom defined by equation (2.9) is a coherent state. Later
Thomson (2001) proposed to call the propagating coherent state after making stationary
phase approximation (saddle point approximation) to the h-f asymptotic solution integral
of wave equation as “elementary coherent state” even the solution may not have a
Gaussian envelope. Therefore, the term coherent state may have different meanings in the
literature. Here we use the term in the general sense and refer to asymptotic solution of
coherent state (CS) as asymptotic CS. The coherent state transform (CST) is defined as
p,x u x x exp 1
2 x 2
eip x d x , (2.31)
where u is the wavefield in space domain, p is the horizontal slowness (ray parameter)
and Ω is inversely proportional to the beamwidth of the coherent state (width parameter).
The inverse transform is
u x
2
2
p,x dp . (2.32)
The coherent-state transform is a complete but non-orthogonal representation. A
property of the coherent-state transform is that the Heisenberg uncertainty cell is
minimized due to the Gaussian windowing. A coherent state can be considered as sum of
weighted plane waves (wavenumber domain summation) or a Gaussian windowed beam
(Gaussian beam) with certain width. In a smooth media, the h-f asymptotic solution (or
semi-classical solution, as called by Klauder, 1987) can be obtained for the coherent state
propagation. The phase is calculated by solving the complex eikonal equation (Hamilton
Jacobi equation), and the amplitude, by the transport equation. Due to the special inverse
transform defined by equation (2.32), the final wavefield is formed by a summation of
coherent states with different directions (p) at the same spatial location without
contributions from neighboring beams (coherent states). This is different from the
Gaussian beam summation which is a summation over neighboring beams with different
directions. Of course, there are other alternative reconstruction formulas can be defined.
However, the inverse transform defined by equation (2.32) is used in all the relevant
articles. In this sense, asymptotic CS (coherent state) modeling bears more similarity with
the Maslov method than with the Gaussian beam method. The Asymptotic CS method
can be considered as a windowed Maslov method; while the Maslov method (see
Chapman and Drummond, 1982; Chapman, 2004) can be said to be the limiting case of
the Asymptotic CS method when the beamwidth is set to infinity (i.e., Ω = 0). The other
limiting case of the Asymptotic CS method is when Ω approaches to infinity, i.e. when
the beamwidth is close to zero, it becomes the classic ray method. Remember that the
wavelength must be approaching zero faster than the beamwidth in order to apply the h-f
asymptotic solution. From the calculations of the eikonal and transport equations, we see
that there are three parameters to consider for the applicability of the method: beamwidth
27
(1/ Ω), wavefront curvature and wavelength. Wavelength being much smaller than the
beamwidth and wavefront curvature is the prerequisite of the asymptotic solution. Under
the asymptotic regime, different approximations can be applied for the cases of large or
small curvature/beamwidth (Klauder, 1987; Thomson, 2001, 2004).
As for numerical modeling schemes, there are different implementations. Foster and
Huang (1991, 2002) apply directly the inverse transform equation (2.32) to synthesize the
space domain wavefield, which is a summation of a bundle of asymptotic CS solutions.
The other way is apply further a stationary phase approximation (or saddle-point
approximation) to the integral and the final solution is similar to a single CS (Klauder,
1987; Thomson, 2001, 2004). In this way, the asymptotic CS ray tracing is similar to the
classic ray tracing with advantage of avoiding caustics. The beam width parameter Ω
enters only into the amplitude calculation as well as the boundary condition of the ray
tracing. The actual beamwidth is irrelevant to the final results. Therefore, unlike the
Gaussian beam dynamic ray tracing where the beamwidth changes along the propagation
path, the asymptotic CS solution assumes a fixed Ω along the ray.
The other difference of coherent state from the Gaussian beam is the inability of
modeling frequency-dependent wave phenomena. The asymptotic CS solution is a global,
uniform asymptotic solution and can handle all caustics, including the pseudo caustics for
which the Maslov method is invalid. It can also give some approximated results even in
shadow zones. However, the solution losses the frequency-dependent wave information
and can only model primary waves. On the other hand, the Gaussian beam approximates
the wave field in the paraxial region of the ray with a parabolic approximation. In the
original frequency domain formulation, the Gaussian beam method can explicitly
simulate frequency-dependent wave phenomena with certain degree of approximation.
Later, time (complex time) domain versions (Hill, 1990, 2001) were also derived for
seismic imaging with high efficiency but with the loss of frequency-dependent wavefield
information. The asymptotic CS method avoids the ray-centered coordinates and each CS
has an asymptotic solution.
Besides some canonical examples for demonstration purpose in the literature,
Albertin et al. (2001) did some comparison of impulse responses and imaging results
between Maslov, Gaussian beam and the asymptotic CS migration methods for a
complex geological model (the Marmousi model).
2.3.4 Beamlet Propagation in Heterogeneous Media by the Local
Perturbation Approach
In non-smooth heterogeneous media, the h-f asymptotic solution for long range
propagation has very limited applications. Wu et al. (2000) developed a local
perturbation theory for wave propagation in the beamlet domain. The propagation of a
wavefield is not synthesized by the superposition of a collection of globally evolving
beams, but by step-by-step beamlet propagation using propagator matrices. At each step,
the laterally varying velocity profile is decomposed into a background velocity profile
and local perturbations. The decomposition in fact is bi-scale decomposition. The large-
scale component is a piecewise homogeneous medium with the scale of window-width
defined for the spatial localization (see Fig. 2.2d); the small-scale component is the local
28
perturbations with respect to the local reference velocities (see Figure 1e). In comparison,
the standard (global) perturbation scheme has a global background velocity (see Fig.
2.2b) and the perturbations are spreading to all scales (see Fig.2.2c).
In the local perturbation approach, the propagator P is decomposed into a
background propagator 0
P and a perturbation term1
P . The background propagation uses
one of the two approximations: the local homogeneous approximation and the average
slowness approximation. Both belong to the h-f asymptotic solution for small-step
propagation. Since the local perturbations are much weaker than the global perturbations,
the perturbation correction operator can be approximated by a phase-screen correction. It
should be understood that the propagation of large-angle waves is mainly controlled by
the background propagator and can be quite accurate due to the wide operator aperture.
Therefore, the beamlet propagator in the local perturbation approach is a hybrid solver
with the asymptotic method for the large scale background media, and the perturbation
method for the small-scale fluctuations with respect to the background (see Wu et al.
2000, Wu and Chen 2001, 2002a, Wang and Wu, 2002, 2003; Luo and Wu, 2003; Chen
et al. 2006, Wu et al., 2007, for detailed derivation and discussions).
One special feature of the beamlet decomposition and propagation is the availability
of local angle information during the propagation. Of course, the wavefield
decomposition into the local angle domain can be performed by local slant stack or local
Fourier transform (e.g., Xu and Lambare, 1998; Sava and Fomel, 2002; Xie and Wu,
2002) during propagation using other types of wave extrapolator. However, the beamlet
propagator is formulated in the local angle domain and therefore is more efficient to get
the angle-related information. Directional illumination analysis has been developed using
the wavefield information in the local angle domain to study the influence of the
acquisition configuration and overburden structures to the illumination of subsurface
targets with different dip-angles (Wu and Chen, 2002b, 2003, 2006; Xie and Wu, 2002,
2003; Xie et al., 2004, 2006). Based on the illumination analysis, Amplitude correction
theory and methods to compensate the acquisition aperture effects have been also
developed (Wu et al., 2004, Wu and Luo, 2005; Jin et al., 2005; Cao and Wu, 2005,
2008, 2009; Mao and Wu, 2010, 2013; Ren et al., 2011).
2.3.4.1 Beamlet Evolution and Wave Propagation in the Beamlet Domain
Wavefield and propagator decompositions:
Substituting the wavefield decomposition equation into the wave
equation (2.17), the beamlets will also propagate obeying the scalar wave
equation,
2 2 2 2ˆ( , ; )[ / ( , )] ( ) 0n m x z mn
n m
u x z v x z b x . (2.33)
Note that in the above equation, ˆ( , ; )n mu x z is a set of coefficients with z as the
labeling parameter, not a variable. The propagation effect of the wavefield now is
29
included in the evolution of beamlets. For a local beam evolution problem, invoking the
one-way wave approximation (neglecting interactions between the forward-scattered and
backscattered waves), we can write a formal solution for the evolution of beamlets
niA zmn mna x e b x
. (2.34)
Where amn is a function evolved from a beamlet bmn propagating in the heterogeneous
medium, and An is the square-root operator
2 2 2/ ( , )n xA v x z . (2.35)
As we mentioned above, amn is no longer a beamlet due to distortion after
propagation. Decomposing amn with the same beamlet basis functions (or with the dual
frame atom in the case of frame beamlet)
,mn mn jl jl
l j
a x a b b x . (2.36)
The propagator matrix P̂ in beamlet domain is as we defined in (2.22) and (2.23)
,ˆ ˆ , ; , ,jl mn l j n m mn jlx x a x b x . (2.37)
The beamlet domain wavefield at depth z z can be obtained as
,ˆ ˆˆ ˆ ˆ( , ; ) ( , ; , ) ( , ; ) ( , ; )l j l j n m n m jl mn n m
n m n m
u x z z P x x u x z P u x z . (2.38)
Here ,ˆ
jl mnP are the matrix elements of the beamlet propagator matrix P̂ , which governs
the beamlet propagation and cross-coupling.
Wavefield reconstruction:
Here we perform the beamlet decomposition using orthogonal bases so the
reconstruction atoms are the same as the decomposition atoms. The wavefield at depth
z z after extrapolation can be reconstructed from the beamlet domain wavefield
through
ˆ ˆ( , ) , ; ( , ; ) ( )n m mn l j jl
n m l j
u x z z u x z a x u x z z b x (2.39)
2.3.4.2 Beamlet Propagator with Local Perturbation Approximation
30
The main task for beamlet imaging (migration) is to derive efficient propagators in
the chosen beamlet domain. The evolution of beamlets is governed by an operator
equation (2.35) which involves a square-root operator and there is no exact solution
available for a general problem. Various approximations are invoked to make the
calculation practical. Exploring the efficiency of fast wavelet transforms and the
sparseness of the propagator matrix, Wu et al. (2000) applied a local perturbation
approximation to the beamlet propagator resulting in a split-step implementation of wave
propagation in the beamlet domain. The local perturbation approach can keep all the
wave phenomena of forward propagation, such as diffraction, interference, scattering and
cross-coupling between beamlets, in the propagator. One successful example of applying
the local perturbation theory is the LCB (local cosine basis) beamlet propagator and
imaging algorithm.
In the local perturbation theory, a local reference velocity 0( , )nv x z is selected for
each window nx , and the local perturbation is calculated from the local reference
velocity. Due to the adaptability of local reference velocities to the lateral variations of
velocity model, generally the local perturbations are small, so that the first order
approximation, i.e. the phase-screen approximation can be adopted for the perturbation-
correction in each window. This leads to the approximation of the square-root operator as
(see Wu et al., 2000; Chen et al., 2006)
2 2 2 2 2 20/ ( , ) / ( , ) ( )n x x n nA v x z v x z k x (2.40)
where 0( ) (1/ ( , ) 1/ ( , ))n nk x v x z v x z denotes the local perturbations.
Therefore, the beamlet evolution equation can be approximated by
1
2n n
i k x z i zi xmn mna x e d e e b
(2.41)
where
i xmn mnb dxe b x (2.42)
is the basis vector in the wavenumber domain, ξ is the horizontal wavenumber and
2 2 20/ ( , )n nv x z (2.43)
is the vertical wavenumber with the local reference velocity. Equation (2.41) is a dual-
domain implementation of an operator split-step approximation. The first factor in the
right-hand side is a phase-screen term in the space-domain; the second factor is a phase-
shift in the wavenumber domain using the local reference velocity and localized by a
beamlet projection. Efficient algorithms of propagation and imaging in the beamlet
domain based on equation (2.41) have been developed using frame or basis beamlets (Wu
and Chen, 2001, 2002; Wang and Wu, 2002; Wu et al, 2003, 2008; Chen et al., 2006).
31
Fig. 2.10 gives a schematic illustration of the decomposition of a lateral velocity
section into a background velocity profile and local perturbations. We see that the
decomposition in fact is a bi-scale decomposition. The large-scale component is a
piecewise homogeneous medium with the scale of window-width; the small-scale
component is the local perturbations with respect to the local reference velocities. In
comparison, the standard perturbation scheme has a global background velocity and
global perturbations spreading to all scales (see the example of SEG salt model in Figure
2.2). We see that the local perturbations are in general much smaller than the global
perturbations, and therefore can reach higher accuracy in extrapolation and imaging in
strongly heterogeneous media.
Fig. 2.10 Background velocity profile (using local reference velocities) and local perturbations in the local
perturbation theory, compared with the global reference velocity (dashed line) and global perturbations in standard perturbation methods.
2.3.4.3. Beamlet Imaging in Strongly Heterogeneous Media
As a numerical example, Figure 3.6 shows the image of the prestack depth migration
using the LCB (local cosine basis) beamlet propagator for the 2D SEG/EAGE salt model.
The minimum velocity of the model is 5000 feet/s and the maximum velocity is
14700feet/s. The salt boundary is sharp and very irregular, especially on the top.
Therefore the model is a strong-contrast and rough heterogeneous medium and h-f
asymptotic method alone will not work. The acquisition system of this model consists of
325 shots (sources) with 176 left-hand-side receivers with receiver interval of 40 feet.
Fig. 2.11a is image using LCB beamlet propagator without acquisition aperture
correction. The aperture-effect corrected image is shown in Fig. 2.11b. We see the
32
excellent quality of the image and the significant improvement of the acquisition-aperture
compensation in the local angle domain.
Fig. 2.11 Images of prestack depth migration using LCB (local cosine basis) beamlet propagator for the 2D
SEG/EAGE salt model (see Fig. 2.1): (a) the image before the acquisition aperture correction, (b) image after
the acquisition-aperture correction. (From Cao and Wu, 2006).
2.4 Curvelet and Wave Propagation
2.4.1 Curvelet and its Generalization
Curvelet transform is originally developed for efficient representation of images with
sharp and curved edges (Candès and Donoho, 2000, 2005). Curvelets are elementary
oscillatory patterns that are highly anisotropic at fine scales, with effective support
obeying the parabolic scaling: width length2. Later the scaling law was loosened and
modified to width wavelength2
in order to let its cousin “wave atom’ and other family
members to live in the same kingdom (Demanet and Ying, 2006). The latter will be
33
discussed later in this section. Fig. 2.11 gives the definition of curvelets and the original
meaning of parabolic scaling. A curvelet is a directional wavelet, indexed by three
parameters: a scale a, 0 < a < 1; an orientation θ, / 2, / 2 and a location b, b ∈
R2. At scale a, the family of curvelets is generated by translation and rotation of a basic
element a
, , ( ) ( ( ))a b ax R x b (2.44)
Here, ( )a x is some kind of directional wavelet with spatial width ~a (perpendicular to
the oscillatory direction) and spatial length (along the oscillatory direction) ~ a :
1/ 0
( ) ( ), ;0 1/
a a a
ax D x D
a
(2.45)
where aD is the parabolic scaling matrix, R is a rotation operator. The essential spatial
support is plotted in the right-hand side of Fig. 2.12. Each flat disk specifies the effective
support of a curvelet in space domain. For two-dimensional patterns or wavefields, the
short axis is along the oscillatory direction and the long axis is perpendicular to that
direction. In beam wave language, the long axis is the effective beam width and the short
axis is along the propagation direction. Showing in the left-hand side of Fig. 2.12 is the
construction of curvelets in the frequency domain (for rigorous definition, see Candès
and Demanet, 2005).
The sampling in the frequency plane (polar coordinates) is also called second dyadic
decomposition (SDD). In the frequency domain, curvelets are supported near a
“parabolic” wedge. The shaded area represents such a generic wedge. Note the duality of
the frequency-domain and space-domain representations. A wide aperture (curvelet
width) corresponds to a narrow wavenumber band, i.e. a narrow lobe of direction, and a
short length in space domain corresponds to a wide frequency-band in the dual domain.
Physically, the parabolic scaling principle for curvelets or the generalized one for its
family represents the requirement of optimum beam-aperture for h-f beam-wave
propagation, which reaches an optimum compromise between the wave spreading and
wave front distortion. A wide aperture, which should contain a few wavelengths, is
necessary for better focusing in order to reduce wide-angle wave spreading; however, the
aperture should not be too large to avoid severe wave front distortion due to
heterogeneities. A good compromise will result in sparse beam (curvelet) propagator
matrix. The beams represented by curvelets may be viewed as coherent waveforms with
enough frequency (wavenumber) localization so that they behave like waves but at the
same time, with enough spatial localization so that they simultaneously behave like
particles. This gives the special advantages of curvelet transform for solving wave
equation in smooth media using asymptotic solutions.
It has been proved that the matrix representation of the Green's function in time-
space domain under curvelet transform is sparse in the sense that the matrix entries decay
34
nearly exponentially fast (i.e., faster than any negative polynomial), and well organized in
the sense that the very few nonnegligible entries occur near a few shifted diagonals,
whose location is predicted by geometrical optics.
Fig. 2.12 Curvelet tiling of phase-space. The figure on the left represents the sampling in the frequency plane,
also called second dyadic decomposition (SDD). In the frequency domain, curvelets are supported near a
“parabolic” wedge. The shaded area represents such a generic wedge. The figure on the right schematically
represents the spatial Cartesian grid associated with a given scale and orientation. (From Candès and Demanet,
2005)
Fig. 2.13 shows the generalization of the curvelets to a family of directional
wavelets. Since they are constructed as tight-frames, they are a family of tight-frame
directional wavelets. From the analysis of optimum beam-aperture for sparse propagator
matrix in smooth media, we can see the extension of curvelet to a family with same
optimum beam-aperture but flexible length requirement (in the oscillatory direction) is a
natural generalization. The extension is realized by generalizing the scaling 2j vs. 2
j/2 to
2αj
vs. 2βj/2
in frequency plane or 2-αj
vs. 2-βj/2
in space plane. From Figure 2.13 we see that
the scaling parameters α and β control the flatness of the support in space, or the
thickness of the wedge in the frequency localization. It is shown that the wave packet
families defined by the horizontal segment at β =1/2 (Fig. 2.12) satisfy the scaling
requirement to yield sparse decompositions of Fourier Integral Operators (Demanet and
Ying, 2006). Along the segment of β =1/2, α =1 corresponds to curvelet, and α =1/2 is the
case of wave atoms. All the transforms specified on the segment can be realized as tight
frames. The other scaling parameters for Gabor atoms, wavelets and ridgelets are also
marked in the figure.
35
Fig. 2.13 Identification of various transforms as (α, β) families of wave packets. The horizontal segment at
β=1/2 indicates the only wave packet families that yield sparse decompositions of Fourier Integral Operators.
(From Demanet and Ying, 2006).
Fig. 2.14 shows the comparison between a curvelet and a wave atom. We can see
clearly the different scaling rules for these two atoms: curvelets obey width = length2 and
wave atoms obey width = wavelength2. Note that for the case of wave atoms, the support
is isotropic. However, for general directional wavelets, the supports are not necessarily
isotropic.
Fig. 2.14 A curvelet (left) and a wave atom (right). Curvelets obey width = length2 and wave atoms obey width
= wavelength2. Note that for the case of wave atoms, the support is isotropic. However, for general directional wavelets, the supports are not necessarily isotropic. (From Demanet and Yin., 2006).
36
2.4.2 Fast digital Transforms for Curvelets and Wave Atoms
Efficient tight-frame constructions and fast digital transform of order O(N2logN) for
curvelets and wave atoms have been developed (Candès et al., 2005; Ying et al., 2006).
As we mentioned before, with tight-frame representation, the decomposition and
reconstruction atoms are the same, and the signal energy is conserved in the frame
domain. In curvelet and wave-atom representations with the fast transform, the
redundancy is 2, i.e. the curvelet or wave-atom coefficients are twice as many as those by
orthonormal transforms.
The construction of the transform can be summarized as (see also Demanet 2006,
§1.3; Chauris, 2006): 1. Fast Fourier transform (FFT) (2D or 3D) the image into
wavenumber domain; 2. For all the scales (1 → N) perform wavenumber filtering
(windowing) using the curvelet function, and inverse FFT the filtered image to get the
curvelet coefficients. For reconstruction, just sum up the contributions from all the
curvelets in the Fourier domain and then transform back to get the space-domain results.
2.4.3 Wave Propagation in Curvelet Domain and the Application to
Seismic Imaging
It has been proved that the matrix representation of the Green's function to wave
equations in time-space domain, or more generally to hyperbolic system of differential
equations, under curvelet transform is sparse in the sense that the matrix entries decay
nearly exponentially fast (i.e., faster than any negative polynomial), and well organized in
the sense that the very few nonnegligible entries occur near a few shifted diagonals,
whose location is predicted by geometrical optics. (See Demanet 2006, §1.1.1 §, 1.2.2;
Candes and Demanet, 2003).
Curvelet transform has been explored and applied to seismic imaging due to its
sparseness in representing multi-dimensional wavefield data and asymptotic propagators
in smooth media (Herrmann, 2003; Douma and de Hoop, 2004, 2005, 2006; Chauris,
2006). The sparseness of seismic data in the curvelet domain is well understood since it is
very similar to the case of image compression. To demonstrate numerically the
sparseness of one-way wave propagators, Douma and de Hoop (2004) gives an example
of curvelet spreading in 2D Kirchhoff migration (Fig. 2.15). In the figure, a curvelet with
dominant frequency 30 Hz (top-left) with coefficients on the spatial lattice (top-middle),
and its spectrum (top-right), is selected to represent an elementary function in space-time
domain excited on the surface. A 2D Kirchhoff migration operator, similar to equation
(2.18) or (2.19), is applied to the curvelet. For a point source in the space domain (a delta
function), the migration (backpropagation) results should be the point spreading function,
which is an elliptic isochrones in this common-offset (fixed source-receiver distance)
imaging configuration. In contrast, the curvelet spreading is rather limited only to its
neighboring cells both in the space and spectral domains (see the middle and right panels
of the bottom row of the figure). This is due to the coherent beam-forming of the spatial
aperture and the time localization of the wide frequency-band signal.
37
Fig. 2.15 The top row shows a curvelet with a dominant frequency of about 30 Hz (left, shown in depth,
2z vt ), the normalized absolute values of the coefficients on the spatial lattice (middle), and its
spectrum (right). The bottom row shows the common-offset Kirchhoff migration of the curvelet in the top row.
The middle panel in this row shows the coefficients on the spatial lattice in the lower left quadrant of the left panel (indicated with the dotted lines) for each of the numbered wedges (labeled `1' to `4') in the spectrum
(right). The middle and right panels of the bottom row show the spreading of a curvelet in space and spectrum,
respectively, during one-way propagation in homogeneous media. Here kx and kz are the horizontal and vertical
wavenumbers, respectively.(from Douma and de Hoop, 2004).
The simplest version of curvelet migration is to perform seismic map migration (For
the original definition and methods of map migration, see Weber 1955, Kleyn, 1977).
Seismic data is first transformed into frequency-wavenumber domain, and then a curvelet
transform is conducted to get the curvelet coefficients. Each curvelet has an initial
propagation direction and location (on the surface) determined by its parameters (x, p)
and the new direction and location after back propagation will be obtained by a rotation
and a time (or space) shift according to the solution of eikonal equation (Douma and de
Hoop, 2004). This is a total geometric mapping. Later a dilation (or stretch) operator is
added to the operation (Douma and de Hoop, 2005; Chauris, 2006) and the imaging
process is called TRD (translation, rotation and dilation) transform. The Algorithm of
Douma and de Hoop (2004, 2005) uses only the largest coefficients after the transform
(migration), resulting in a very efficient migration process (two orders of magnitude fast
than the ray-Kirchhoff migration). Chauris (2006) showed that interpolation (in his case,
the Shannon interpolation scheme) is necessary to keep the appropriate accuracy. He
showed an example of time migration for complex reflection structure taken from the
Marmousi model but with a uniform background. The data were regenerated by a ray-
Born method with the original reflectivity embedded in a homogeneous medium. The
imaging result is as good as the Kirchhoff migration. In Douma and de Hoop (2006), an
38
extra dilation term was introduced based on a stationary phase approximation to the
asymptotic solution to account for the focusing/defocusing effects and the geometric
spreading of curvelets. Numerical tests for simple models in homogeneous background
show the similarity of its image quality to that of the ray-Kirchhoff migration.
2.5 Wave Packet: Dreamlets and Gaussian Packets
2.5.1 Physical wavelet and wave-packets
Space-domain wavefields or seismic data recorded on a surface are special data sets.
They are not arbitrary pictures but rather the results of wave propagation and scattering.
They observe certain rules and cannot fill the 4-D space-time in arbitrary ways. The time-
space distributions must satisfy the causality relation which is dictated by the wave
equation. In the scalar case the wave equation is
(2.46)
The solution of the wave equation in fact is very sparse in the 4D volume. It can
only occupy a hyper-surface. For example, in the case of homogeneous media, the wave
solution can only exist on the “light cone” in the 4D space-time (Fig. 2.16a) or its
Fourier space (Fig. 2.16b). There are immense amount of points in the 4D
continuum outside the light cone that is excluded from the wave solutions because they
do not satisfy the causality relation (they propagate too fast or too slow). For a
inhomogeneous media with smoothly varying velocity, we can attach a local “light cone”
to each point, and the causality relation has to be satisfied locally at each point. The “light
cone” is a critical constraint to effectively represent wavefields. That is why physical
wavelet can greatly sparsify seismic and other wavefield data.
The light cone along the positive time-axis is the “causal cone”, and the one along
the negative time-axis is the “anti-causal cone”. Regular seismic data are situated on the
causal cone; Only backpropagated or time-reversal wavefields are located on the anti-
causal cone.
“Physical wavelet” was introduced by Kaiser (1993, 1994, and 2003) as localized
wave solutions to the wave equation by extending the solution to the complex space-time.
By introducing properly the imaginary-axis in time or space, wave solutions can be
localized, i.e. only exist around a space-time point and attenuating very quickly when
away from the central space-time point; But these wave-packets still satisfy the wave
equation and can be used as elementary wavefields (wavefield atoms) to synthesize any
complicated wavefields. In fact, the idea of localizing wave solution by complex
extension has been discussed by many authors (Felsen, 1976; Enziger and Raz, 1987;
Heyman and Steinberg, 1987). Kaiser systematically developed the theory and termed it
as “physical wavelet”. The wavelets derived in this way not only possess the properties of
the wavelet, but also satisfy automatically the wave equation, which is a distinctive
feature different from mathematical wavelets. This feature is very desirable when applied
to physical problems such as wave propagation and imaging.
2 2( ) ( , ) 0t u t x
( , )u tx
( , )tx
( , )p
39
Fig. 2.16 The light cones: (a) The causality hyper-surface in the 4D space-time; (b) The dispersion hyper-
surface in the 4D Fourier space.
A wave solution can be expressed in a form of plane wave superposition in the 4D
Fourier domain. In order to express the wave solution as a superposition of localized
waves, Kaiser (e.g. 1994) applied an AST (analytic signal transform) to , resulting in
an extension of from the light cone in real space-time to a causal tube in
complex space-time
(2.47)
where is the integration weight (Lorentz-invariant measure) on the
light cone, is the 3D wavenumber vector, and
(2.48)
is an acoustic wavelet of order (Kaiser, 1994), with denoting the unit step
function. In equation (2.47) the inner product of a wavefield with a physical wavelet
(2.48) is the decomposition of wavefield into localized solutions, the physical wavelets.
The localization is realized by the windowing process in the 4D Fourier domain. The
window function is defined in the causal tube. Points in the causal tube have projections
on the light cone in the form of windowing. From equation (2.48) we see that the pulse
width (waveform) is controlled by and the beamwidth-steering is parameterized by
. In Fig. 2.17 we show the wavepacket (physical wavelet) on the light
( )u x
( )u x ( , )x t x
( )u x iy
*
1( ) ( , ) ( , )zC
dpu x iy u
k
p p
3 3( /16 )dp d c p
p
1( , ) 2 ( ))exp ( ) ( )z k i t p p y p x p y
(.)
1 2 3( , )y y yy
40
cone as yellow spots. Since we take only the causal solution (for positive time), there
exist a pair of yellow spots on the light cone in the Fourier space. The windowing on the
light cone will have projected windows in the frequency-axis and the wavenumber plane.
Fig. 2.18 shows some examples of the time-domain waveform (for a fixed space location)
of different orders ( =3, 10, 15, 50) of a physical wavelet. In the same way, we show
the space localization at different times in Fig. 2.19 for the same physical wavelet. In this
example, the wavelets are isotropic ones, so the plot is for the r-dependent.
Fig. 2.17 Windowing (localization) on the light cone in the Fourier space ( ), where p is the space
wavenumber vector and is the frequency.
Fig. 2.18 Physical wavelets along the time-axis (at r = 0) for . Solid lines are the real part,
and dotted lines are the imaginary part.
,p
3,10,15,50
41
Fig. 2.19 Physical wavelets along the space-axis (r-axis) (at t = 0) for isotropic wavelets at .
2.5.2. Dreamlet as a Type of Physical Wavelet
A space-time physical wavelet is a localized solution of the corresponding wave
equation. As we discussed in the previous section, it can be formed by analytical
extension to a complex space-time domain, so the imaginary part of the wavefield
controls the localization of the solution. In fact, there are other ways to obtain localized
solutions of wave equation, such as the Gaussian packet introduced by Kiselev and Perel
(1999) and Perel and Sidorenko (2007).
Wu et al. introduced a space-time wavelet formed by a tensor product of wavelet in
time-domain (drumbeat) and beamlet in space domain, and termed it as dreamlet
(drumbeat beamlet) (Wu et al., 2008, 2009; Wu and Wu, 2010). The relation between
the dreamlet and physical wavelet is discussed theoretically in Wu et al. (2011). Here we
explain conceptually the relation between physical wavelet, dreamlet, and other wavelets
defined in a subspace of the 4D space-time.
Time Slice and Depth Slice of 4D Wavefield:
In Fig. 2.20 and Fig. 2.21 we show the different features of “time-slice” (snapshot)
and “depth-slice” (data section) of a 4D wavefield. We see that a “snapshot” is the
wavefield distribution in the 3D space domain for a given time t. Due to the causality, the
solution can only exist in a hollow sphere (in 2D, a circle) (Fig. 2.20); While for a depth-
slice (data on a plane), it becomes a “paraboloid” (Fig. 2.21). They appear to have quite
different picture structures, but are in fact from the same object. Therefore, the
3,10,15,50
42
x z
decomposition atoms for efficiently representing a wavefield (snapshot) or a data set
(seismograms) could be very different. For the snapshots in the space-domain, local
Fourier atom, Gabor atom, curvelet, or other directional wavelet can be used due to the
symmetry property of the time-slice; On the other hand, the data decomposition (depth-
slice) need different atoms to take into account of the paraboloidal structure. “Dreamlet”
and different wavepacket decomposition are designed for this purpose.
Fig. 2.20 A time-slice (snapshot) cut through a light cone.
Fig. 2.21 A depth-slice (data section) cut through a light cone.
43
Fig. 2.22 Examples of time-slices (snapshots) of the wavefield from the 2D acoustic SEG/EAGE salt model. The source is an 18 Hz Ricker wavelet.
Fig. 2.23 Examples of depth slice of the wavefield: the seismic data (acquired on the earth suface) of SEG 2D
salt model (coincided source and receirver)
In Fig. 2.22 we show two snapshots of the propagated wavefields for the 2D acoustic
SEG/EAGE salt model. The source is an 18 Hz Ricker wavelet. We see that the
wavefields are basically composed of different circles of various radii (the curvature of
the circle changes with the propagation velocity). In contrast, the seismic data set
Distance (km)
44
acquired on the earth surface (depth-slice) shown in Fig. 2.23 (post-stack data) for the 2D
acoustic SEG/EAGE salt model, resembles a collections of parabolas of different sizes.
This is consistent with the concepts demonstrated in Fig. 2.20 and Fig. 2.21.
Dreamlet =DrumbeatBeamlet:
A dreamlet atom is in the form
( , ) ( ) ( )tt x xd x t g t b x (2.49)
where
i ttg t W t t e
(2.50)
is a t f atom (drumbeat) with W t as a smooth window function, and
( ) ( ) i xx
b x B x x e (2.51)
is a x atom (beamlet) with ( )B x as a bell function. In this section, x represents a
horizontal position on the observation plane. The bars over letters signify the variables
are local variables related to the window centers. Therefore, ( , , , )t x marks the
coordinates of local time, local frequency, local distance, and local wavenumber,
respectively. Note that the phase terms in the t f atom and the x atom have
opposite signs. This is consistent with the causality relation imposed by the wave
equation. Although the beamlet transform is applied on the observation plane, however,
the beamlets are defined in the full wavenumber space through the dispersion relation.
The window ( )B x x on the observation plane (here the x-axis for the 2D case) is a
cross-section of the whole space window.
Note that in the definition (2.50) and (2.51), the phase terms are still using global
space-time. An alternative definition of the dreamlet atom is to use the local space time:
( )
( )( ) ( )
i t tt
i x xx
g t W t t e
b x B x x e
(2.52)
This definition is more consistent with the definition of the local cosine bases. It is only a
matter of phase-shift of the atoms, and all localization properties are kept same.
2.5.3. Seismic Data Decomposition and Imaging/Migration using
Dreamlets
45
2.5.3.1. Seismic Data Compression and Data Recovery in Compressive
Sensing using Dreamlets:
A space-time domain seismic data field or wavefield on an extrapolation plane can
be represented as superposition of the phase space atoms (dreamlets)
( , ) ( , )t x t x
t x
u x t u d x t
(2.53)
where t x
u are the dreamlet coefficients. As we discussed in previous sections, seismic
data are not ordinary pictures, but are produced by physical process satisfying the wave
equation. They can only situate on the depth-slices of the “light cone” in the 4D time-
space. Therefore, the representation by physical wavelets is perhaps the sparsest one. In
the following we show two examples demonstrating the representation efficiency and
sparseness of dreamlets in comparison with out methods, such as curvelets. Fig. 2.24
shows the compression ratios of the SEG 2D salt model data ( see Fig. 2.23) by different
decomposition methods (including the beamlet and curvelet) as a function of threshold
(Wu et al., 2008; Geng et al., 2009). We see that Dreamlet, or physical wavelet in
general, has the most efficient representation to this seismic data set, especially for high
compression ratio.
Fig. 2.24 Compression ratios of different decomposition methods for the SEG/EAGE salt model poststack data.
The second example is on the data recovery in compressive sensing (Donoho, 2006;
Herrmann, 2010). The compressive-sensed data set with 50% missing traces is simulated
by removing half of the geophone records randomly in the post-stack data set of
46
SEG/EAGE 2D salt model. Data recovery (recovery of the missing traces) is done by a
basis pursuit method based on l1-norm optimization to search for the most efficient
representation (least residual with minimum coefficients). As shown in Fig. 2.25,
Dreamlet method takes around 60 iterations to get the solution while Curvelet method
takes around 950 iterations with similar recovery quality (signal-to-noise ratio) (Wu et
al., 2013). From the Figure we see also that dreamlet has less number of coefficients to
represent the recovered data.
Fig. 2.25 Data recovery in compressive sensing by dreamlet and curvelet: L2-norm of the residuals (top panel)
and number of coefficients (bottom panel) as function of iteration number.
2.5.3.2 Dreamlet evolution
The evolution of dreamlets in heterogeneous media can be derived from the wave
equation. Assuming a wave propagator, here a one-way wave propagator, , is applied
to a dreamlet, the propagator matrix in the dreamlet domain is obtained as
t x
t x t x t xd d
(2.54)
where f g stands for the inner product of f and g, t x
d is the dreamlet atom at the
current level (depth), and t x
d is the dreamlet at the next level during propagation. d
is the dual frame vector of d. As in the case of beamlet propagator (see section 2.3.4), a
dreamlet propagator can be derived by the local perturbation theory. The use of local
background velocities and local perturbations results in a two-scale decomposition of the
dreamlet propagator: a background propagator for large-scale structures and a local
phase-screen correction for small-scale local perturbations (for detailed derivation, see
Wu et al., 2008, 2009; Wu and Wu, 2010).
47
Fig. 2.26 shows some examples of the Gabor dreamlet atoms and their snapshots
during propagation. Gabor atom is characterized by the use of Gaussian window and
exponential phase function. In the example, both the space and time windows are of L=32
points Gaussian window. Compared with the beamlet propagation, we see that the
dreamlet propagation maintains both the space and time localizations in homogeneous or
smoothly inhomogeneous media. For strongly heterogeneous media, the dreamlet will
split and spread due to diffraction and scattering. Similar to the beamlet propagation, we
apply the local perturbation method to calculate the propagation effect of small-scale
perturbations in each window (under one-way approximation).
We can also use orthogonal bases, such as the LCB (local cosine basis) as the
dreamlet atoms. Both decomposition and propagation using LCB dreamlets are more
efficient due to the orthogonality of representation. However, similar to the LCB beamlet
propagation, the LCB dreamlet has always two symmetric lobes.
Fig. 2.26 Gabor dreamlet atoms (space-time representation) (Top panels) and their snapshots (for 3 time-
instances) during propagation (Bottom panels). Window lengths for both time and space are Gaussian windows
with L=32 points. We see that atoms with different space-time patterns propagate in different localized directions.
2.5.3.3 Application of Dreamlet to Seismic Imaging
Dreamlet migration/imaging is composed of three major steps:
1. Data decomposition: On the surface, the data set is decomposed into dreamlets.
48
2. Wavefield extrapolation (downward propagation): in the dreamlet domain: For each
step z , apply the propagator matrix to the coefficient and obtain a new set of
coefficient. This step include both the propagation in local homogeneous reference
media and the local phase-screen corrections.
3. Apply the imaging condition at each step to obtain the image field: This can be done
either in the dreamlet domain, or by transforming back to the space-time domain by
the inverse dreamlet transform. Step 2 and 3 are implemented iteratively until the
bottom of the image space.
Fig. 2.27 Number variation of dreamlet coefficients during migration. The black and green lines are for the
survey sinking dreamlet coefficients using sunken data and the full data, respectively (See Wu et al., 2013).
The salient feature of the dreamlet migration is the ability to implement
“migration/imaging in compressed domain”. Due to the high efficiency in representing
the seismic data and wavefield, the dreamlet coefficients becomes very sparse, and can
stay sparse during imaging process such as the prestack depth migration. From Fig. 2.25,
we see that in general, the dreamlet coefficients will stay sparse during
migration/imaging process. The example shown here is for case of survey-sinking
migration, in which both the source array and receiver array are downward-extrapolated
step by step in depth, and apply the imaging condition locally at each depth (see, Wu et
al.,2009, 2013). After sinking the surveying system to the depth, imaging process
(determining the scattering points) is realized locally by coinciding incident and scattered
waves at the imaging points (zero-traveltime imaging condition). The green line shows
the variation of coefficients during migration using the full data, which is similar to the
case of shot-domain prestack depth migration; while the black line shows the variation
when using the sunken data during survey sinking imaging. Because of the time
localization in dreamlet decomposition, the data with negative traveltime which will have
no contribution to the imaging process beneath the current depth will be discarded. In this
way, the already sparse coefficients become sparser during the migration process (Fig.
49
2.27). Sparse data representation, recovery, imaging and inversion form an active area of
research.
2.5.4 Gaussian Packet Migration and Paraxial Approximation of
Dreamlet
When we use the Gabor frame atom (also called Gabor-Daubechies frame) as both
the drumbeat (t-f atom) and the beamlet atoms ( x atom) to form the dreamlet, the
atom is called Gabor dreamlet (section 2.5.3.2). Since the window in both time and space
is of Gaussian shape, it is also called Gaussian wave packet or simply Gaussian packet
(Norris et al.,1987; Klimeš, 1989; Kiselev and Perel, 1999; Červený, 2001; Perel and
Sidorenko, 2007; Qian and Ying, 2010). In case of smooth media, h-f asymptotic solution
exists for the Gaussian packet and the packet will travel along the central ray determined
by the eikonal equation (see, Červený et al., 2007). The asymptotic solution of
wavepacket is also called “quasiphotons (Babich and Ulin, 1981), “space-time Gaussian
beam” (Ralston, 1983), “coherent state” (Klauder, 1987).
Dreamlet evolution in arbitrarily heterogeneous media can be calculated by dreamlet
propagator matrix (2.54). However, in smooth media, the scattering is weak, and
dreamlet evolution only involves shift, rotation, wavefront spreading, envelope
broadening and other distortions. These forward-scattering phenomena can be calculated
approximately by h-f asymptotic method. In this approach, once the data set is
decomposed on the surface, the wavepacket (dreamlets) can propagate globally and
individually without mutual coupling. This method of asymptotic solution is convenient
and efficient, though less accurate. It has great potential in many applications.
High-Frequency Asymptotic Solution and Paraxial Approximation:
Traditional high-asymptotic solution (e.g., Červený et al., 2007) can be applied to the
dreamlet evolution
( , , ) , , exp , ,t x
d x z t A x z t i x z t (2.55)
where can be understood as the central frequency of the wave packet, and both phase
function ( , , )x z t and amplitude function ( , , )A x z t are complex-valued functions of
space coordinates ( , )x z and time t . For simplicity, we use instead of ( , , )x z t and
A instead of ( , , )A x z t in the following equations. From the definition of (2.49) –
(2.51), we see that dreamlet atoms with Gaussian windows can be treated as Gaussian
wavepacket. In the following we summarize the method of paraxial Gaussian packet and
establish the link of seismic data decomposition using Gaussian packet to the dreamlet
decomposition.
For the convenience of derivation and efficiency of propagation algorithm, the
accurate complex phase function (traveltime field) of a wavepacket can be expanded into
second order of Taylor series in the vicinity of its center along a ray. Under this
50
approximation, not only the amplitude envelope, but also the phase front become
Gaussian (parabolic approximation). That is the origin of the name of Gaussian beam and
Gaussian packet. Červený (2007) call this approximation as “paraxial Gaussian packet”,
and call the original asymptotic solution as “strict” Gaussian beam or Gaussian packet.
The propagation of the paraxial Gaussian packet using the asymptotic solution is well
documented in the literature (e.g. Klimeš, 1989; Červený, 2007). Here we summarize
some major points and explain the results in terms of its wave physics
Under the h-f asymptotic approximation, the wavepacket will move along a ray
satisfying the eikonal equation
2 2( ) 0v
t t
(2.56)
where v is the local velocity. The eikonal equation determines the trajectory of the
wavepacket movement, such as its shift and rotation. In fact, the eikonal equation is the
mathematical form of saying that the trajectory must stay on the local light cone, and is a
form of causality relation (dispersion relation) of wave physics. On the other hand, to
describe the distortion of the wavepacket, the phase function was expanded into in a
second order Taylor expansion in the vicinity of its central point along the ray,
21
( ) ( , ) ( ) ( )2
r rt x x p x x N x x (2.57)
where x is the 4-dimensional space vectors, ( , / )t p is 4-dimensional
slowness vector (gradient of the phase function), and N is the 4-dimensional curvature
matrix (second derivative of the phase function), respectively. rx is the central point of
the packet along the ray. For the 2D case, equation (2.57) can be written explicitly
2 2 22 2
2
( , ) ( ) ( )
1 1( ) ( ) ( )( ) ( )( )
2 2
r r
r r r r r r
t t tt
t t x x t t z z t tt x t zt
x p x x
N x x
(2.58)
where p , x and rx are 2D vectors (in 3D media: 3D vectors). Note that here p denotes
slowness vector (not to be confused with the case of physical wavelet, where it is
wavenumber vector). Due to the causality relation, it should hold in the above equation
1t
(2.59)
Under this paraxial approximation, equation (2.55) becomes
51
2
( , ) , , , , exp , ,
1exp ( ) ( ) ( )
2
pgp pgp pgpt x
pgpr r r
d t u x z t A x z t i x z t
A i t t
x
p x x N x x
(2.60)
Where pgpu ,
pgp , and pgpA are corresponding functions of “paraxial Gaussian
packet”.
The above equation describes the paraxial Gaussian Packet (PGP) evolution in the
Cartesian coordinate system. Changing to the ray-centered coordinate system ( , , )q s t ,
where q is the perpendicular distance from the central ray and s is the distance along the
central ray. The curvature matrix becomes
2 2 2
2
2 2 2
2
2 2 2
2
qq qs qt
sq ss st
tq ts tt
q s q tqM M M
M M Ms q s ts
M M M
t q t s t
M (2.61)
Then the curvature matrix and other quantities needed for the packet evolution in
equation (2.58) can be efficiently calculated by kinematic and dynamic ray tracings (see
Červený, 2007). Therefore, the main calculation task is the dynamic ray tracing for P
(slowness derivative) and Q (position derivative). Once the spatial curvature qqM is
determined by the DRT (dynamic ray-tracing), the temporal curvature and other
quantities can be determined by analytical relations (Klimeš, 1989; Geng et al., 2013).
This is also due to the causality relation which restricts these curvatures from arbitrary
variation. This can explain the efficiency of the wavepacket method compared with the
beam method. In order to calculate the frequency-dependent phenomena of wave
propagation, we need to compute beam propagation for many frequencies in a frequency-
band. The final waveform change is the result of interference between all the beams in
the band. However, for smooth media, this change can be predicted by the parabolic
approximation, which is the formulation of wavepacket propagation.
Wavepacket Decomposition of Seismic Data:
An important issue in the application of Gaussian packet in seismic imaging, the
decomposition of seismic data into Gaussian packet, was not well-solved until recently
and is still an active research area.
In the literature, some effort has been paid to the decomposition of seismic data
recorded on the surface into paraxial Gaussian packets (Zacek 2004, 2005, 2006a,b).
52
However, the proposed method is time-consuming and the imaging results are not
satisfactory. From the discussion on physical wavelet and dreamlet, we know that seismic
data are generated by exact wave equation, not by paraxial wave equation. Therefore, the
decomposition using paraxial Gaussian packets does not fit the task; while dreamlet atom
may be one of the optimum decomposition atoms for seismic data. From Fig. 2.24 and
Fig. 2.25 we see the high efficiency of dreamlet representation due to the property of
physical wavelets. The other reason of the efficient representation of seismic data by
dreamlets is the relatively narrow band of seismic signal in exploration seismology. In
Fig. 2.28 we show the fitting of a Ricker wavelet with central frequency at 15 Hz (solid
line) by a Gabor time atom (strict Gaussian packet) of different parameters. It can be seen
that with 2 2/ 20ttM t i , a single Gaussian packet can well represent the seismic
source time function.
After dreamlet decomposition, data compression can be done with thresholding.
Depending on the data sets, the compression ratio varies. In the migration example of
shown later in Figure 2.31, only 1% of coefficients are used for imaging. The conversion
from dreamlet parameters to the initial parameters of paraxial Gaussian packets has been
discussed (Geng et al., 2013).
Fig. 2.28 Comparison of the time-profile of a Gaussian packet of different parameters (2 2
33 /M t )
with the Ricker source wavelet. We see that with the choice of appropriate parameter (M33=20i), the seismic source wavelet can be simulated by a single Gaussian packet.
Examples of Guassian Packet Evolution, Modeling and Imaging:
Fig. 2.29 shows an example of the evolution of three Gaussian Packets of
20f Hz with different propagating directions in a vertically gradient media, which has
velocity of 2000 /m s at the surface and 2900 /m s at depth of 3000m . Black lines
are the three central rays for the corresponding Gaussian Packets. For comparison, the
evolution results using 2D fourth-order acoustic FD algorithm are shown at the bottom.
We see that the Gaussian Packet solution in this simple medium can provide fairly
accurate phase and amplitude information around the vicinity of the central point.
However, the phase begins to be inaccurate when it is too far from the central point.
53
Fig. 2.29 Evolution of an initial Gaussian packet in a linearly gradient medium. a) by paraxial Gaussian Packet
method; b) by full wave Finite Difference method.
Second example is for the case of complex model. In order to perform ray tracing,
the 2D SEG-EAGE velocity model is smoothed to avoid singular behavier of rays on the
salt boundaries. Even after smoothing, the velocity model is still strongly varying in
irregular ways. Fig. 2.30 shows the impulse response by the Gaussian packet summation
method. We shot rays with the same initial time from the surface for all the Gaussian
Packets centered at different time rt . The wavefield at time T in the subsurface can be
approximated by the summation of all Gaussian Packets traveled to the point ( , )r rx z
with traveltime rT t . The source time function used here is a Ricker wavelet with
dominant frequency 30f Hz and is decomposed into a superposition of Gabor frame
with redundancy 8 and window length of 64. Gaussian Packets are shot from the source
point with 200 directions. The comparison with FD method shows that the Gaussian
packet method can simulate most of the prominent features of the impulse response
outside the salt body or after traveling a short distance inside the salt body. However,
after long propagation inside body, the distortion becomes more severe.
Fig. 2.31 gives an example of Gaussian packet migration for the 2D SEG-EAGE
zero-offset data set. As mentioned earlier, the salt boundaries are smoothed for the sake
of ray-tracing. The minimum velocity is 1600 /m s and the maximum velocity is
4500 /m s . The data decomposition uses the Gabor dreamlet, with frame redundancy 8
and window length of 64 for both time and space. On the top is the image obtained by
beamlet migration for comparison. At the bottom shows the image by the Gaussian
packet migration. Only 1% of the largest dreamlet coefficients are used in the migration.
Compared to LCB (wide-angle one-way migration) method, Gaussian packet method can
54
recover almost all the important structures of the model, such as the boundary of the salt
body, the steep faults and most of the subsalt structures.
Fig. 2.30 Impulses responses (by Gaussian packet summation) in the 2D SEG-EAGE model calculated by the
paraxial Gaussian Packet method at 1.0 t s (b) and at 1.5 t s (d). For comparison plot on the left are
those calculated by full-wave finite-difference method (FD); The source location is 8550 x m .
Fig. 2.31 Comparison of imaging results for SEG-EAGE 2D model. Top: By wide-angle one-way imaging
method (LCB beamlet migration); Bottom: By paraxial Gaussian packet method with Gabor dreamlet
decomposition of redundancy 8 and 64N N .
However, some imaging artifacts appear at the irregular top salt boundary due to the
strong scattering effect. It is also noticeable that in some subsalt area, Gaussian Packet
55
method has weaker images due to the reduced number of effective packets which can
penetrate the thick and irregular salt body. However, Gaussian packet migration is much
faster, and has the flexibility to include large-angle and turning waves in migration.
Therefore, the research along this direction is still active.
2.6 Conclusions
1. For wavefield decomposition, both beamlet and curvelet transforms have elementary
functions of directional wavelets. Beamlet is a type of physical wavelet, representing
an elementary wave (satisfying wave equation) in various wavefield decomposition
schemes using localized building elements, such as coherent state, Gabor atom,
Gabor-Daubechies frame vector, local trigonometric basis function, etc.. Curvelet
transform is a specifically defined mathematical transform, characterized by the
parabolic scaling: width length2. A family of beamlets is built on the invariance
with space shift, dilation and frequency (wavenumber) shift; while curvelets are
based on the invariance with space shift, dilation and rotation. In curvelets the
directivity is directly defined by the rotation parameter; while in the case of beamlets
the directivity is related to the wavenumber by the wavenumber-angle relation (2.7).
2. Sparseness of propagator matrix and the parabolic scaling: The parabolic scaling law
originally is defined as width length2
for the elementary function in the space
domain. Later it is relaxed to width wavelength2
in order to let “wave atom’ and its
other cousins stay in the family. In this way, the length requirement is more flexible.
This extended parabolic scaling law is similar to the beam-aperture requirement for
asymptotic beam solution a > >> , namely the beamwidth must be smaller than
the scale of heterogeneity and much greater than the wavelength. Optimal
beamwidth is reached by balancing the beam geometric spreading which is
controlled by the ratio / , and the beam-front distortion which depends on a /
. Using optimal beamwidth, beamlet or curvelet propagator will be sparse in smooth
media for short range propagation, since the elementary function will propagate with
least distortion and cross-coupling. Even through the parabolic scaling is only a
special case of the general criterion, the efficient tight-frame construction of curvelet
transform based on this scaling is very useful. For long range beam propagation,
there are two more parameters enter into the game: the range R and the total
perturbation strength . The requirement of a >> for asymptotic solution is
replaced by / 2Fa r R where Fr is the Fresnel radius along the beam path
(see section 2.3.3)
3. Asymptotic methods for beam propagation: In smooth media, various asymptotic
theories and methods have been developed for propagation of beams – globally
evolved beamlets: The Gaussian beam, complex ray, pulsed beam, coherent state and
other asymptotic methods. These solutions are either in a form of wavenumber
integral or an explicit space-domain solution. These solutions have very similar
56
derivations, although they may have different degrees of approximation. Asymptotic
solutions for curvelets have also emerged recently.
4. Beamlet scattering in heterogeneous media: In case that the beam-aperture
requirement a > >> is not satisfied, or other criteria of asymptotic analysis are
violated, beamlet or curvelet scattering will occur and needs to be studied. For
beamlet propagation, a local perturbation theory and method have been developed to
handle strong and rough heterogeneities. The propagator is decomposed into a
background propagator and a perturbation operator for each forward marching step.
For background propagation asymptotic solutions can be applied to the piece-wise
background media, and a phase-screen correction is performed for the local
perturbations. Numerical examples demonstrated the superior image quality for
subsalt structures.
5. Dreamlet (Drumbeat Beamlet) is a fully localized wavefield atom, and is a type of
physical wavelet (satisfying the wave equation automatically). It is efficient in
representing seismic data and other wavefield data (with high compression ratio).
Dreamlet propagator is also sparse, and asymptotic approximation leads to a
Gaussian packet propagator which has been applied to seismic migration/imaging.
6. Future work on beamlet/curvelet and dreamlet scattering in generally heterogeneous
media is needed for the development of efficient methods on propagation and
imaging. The multi-scale nature and the advantages of phase-space localization have
to be further studied and put into applications in imaging and inversion.
Acknowledgement
We have benefited from the discussion with Yingcai Zheng, Jun Cao, Xiao-Bi Xie and
Chuck Mosher. Yingcai has contributed to the review of Gaussian beam and coherent
state. Yu Geng, Yaofeng He, Jian Mao, Lingling Wang, Bangyu Wu have participated in
the preparation of the draft and helped in drawing some figures. We are very grateful to
Laurent Demanet, Huub Douma and Martijn de Hoop for discussion and allowing the use
of their figures in this review article. We would like to acknowledge the support from
WTOPI (Wavelet Transform On Propagation and Imaging for seismic exploration)
Project and the DOE/Basic Energy Sciences project at University of California, Santa
Cruz, California, USA and the support from the China NSF and other projects of Xi’an
Jiaotong University.
References
Aki, K. and Richards, P.G., 1980, Quantitative seismology, theory and methods, vol. 2:
W.H. Freeman and Company, San Francisco.
57
Albertin, U., Yingst, D., and Jaramillo, H., 2001, Comparing common-offset Maslov,
Gaussian beam, and coherent state migrations: Expanded abstracts, SEG 71st Annual
Meeting.
Albertin, U., Yingst, D., Jaramillo, H. and Wiggins, W., 2002, Towards a hybrid
raytrace-based beam/wavefield-extrapolated beam migration algorithm: Expanded
abstracts, SEG 72nd Annual Meeting, 1344-1347.
Ali, S.T., Antoine, J-P, and Gazeau J-P, 2000, Coherent states, wavelets and their
generalizations: Springer, Chapter 15.
Antonie, J. –P., and Murenzi, R., 1996, Two-dimensional directional wavelets and he
scale-angle representation, Signal Processing 52, 259-281.
Antonie, J. –P., 2004, The 2-D wavelets transform, physical applications and
generalizations: Wavelet in Physics, edited by J.C. Van den Berg, Cambridge Univ.
Press, 23-75.
Auscher, P., 1994, Remarks on the local Fourier bases, in Wavelets, Mathematics and
applications, edited by J.J. Benedetto and M.W. Frazier: CRC Press, 203-218.
Averbuch, A., Braverman, L., Coifman, R., Israeli, M., Sidi, A., 2002, Efficient
computation of oscillatory integrals via adaptive multiscale local Fourier bases, Appl.
Comput. Harm. Anal. 9-1, 19-53.
Babich, V., and V. V. Ulin, 1984, Complex space-time ray method and 'quasiphotons':
Journal of Mathematical Sciences, 24, 3, 269-273.
Balian, R., 1981, Un principe d’incertitude fort en théorie du signal ou en mécanique
quantique, C . R. Acad. Sci. Paris Sér. II, 292, pp. 1357-1361.
Baraniuk, R.G. and Jones, D.L., 1992, New dimensions in wavelet analysis: Proc. IEEE
intern. Conf. Acoust. Speech Signal Process., IEEE Press, Piscataway, NJ.
Bastiaans, M.J., 1993, Gabor’s signal expansion and its relation to sampling of the
sliding-window spectrum, in Marks II JR (1993), ed., Advanced Topics in Shannon
Sampling and Interpolation Theory, Springer-Verlag, Berlin.
Battle, G., 1988, Heisenberg proof of the Balian-Low theorem, Lett. Math. Phys. 15,175-
177.
Battle, G., 1992, Wavelets: A renormalization group point of view, in Ruskai MB,
Beylkin G, Coifman R, Daubechies I, Mallat S, Meyer Y, and Raphael L, eds.,
Wavelets and their Applications, Jones and Bartlett, Boston.
Benedetto, J.J. and Frazier M.W., 1993, Wavelets: Mathematics and their Applications,
CRC Press, Boca Raton.
Benedetto, J.J. and Walnut, D. F., 1993, Gabor frames for 2L and related spaces, in
Benedetto JJ and Frazier MW, eds., Wavelets: Mathematics and their Applications,
CRC Press, Boca Raton.
Benedetto, J.J. and Zayed, A.I., 2004, Sampling, wavelets, and tomography: Birkhäuser.
Beylkin, G., 1985, Imaging of discontinuities in the inverse scattering problem by
inversion of a causal generalized Radon transform, J. Math. Phys. 26, 99-108.
Beylkin, G., Coifman, R., and Rokhlin, V., 1991, Fast wavelet transforms and numerical
algorithms, Comm. Pure Appl. Math.44, 141-183.
Beylkin, G. and Sandberg, K., 2005, Wave propagation using bases for bandlimited
functions, Wave Motion, 41-3, 263-291.
Candès, E.J., 1999, Harmonic analysis of neural networks. Appl. Comput. Harmon. Anal.
6, 197-218.
58
Candès, E.J. and Demanet, L., 2003, Curvelets and Fourier integral operators. C. R.
Acad. Sci. Paris, Ser. I 336, 395-398.
Candès, E.J. and Demanet, L., 2005, The curvelet representation of wave propagators is
optimally sparse, Comm. Pure Appl. Math. 58-11, 1472-1528.
Candès, E.J., Demanet, L., Donoho. D.L., Ying, L., 2006, Fast Discrete Curvelet
Transforms, SIAM Multiscale Model. Simul., vol. 5-3, pp. 861–899.
Candès, E. and Donoho, D., 1999, Ridgelets: the key to high-dimensional intermittency?
Phil. Trans. R. Soc. Lond. A. 357, 2495-2509.
Candès, E. and Donoho, D., 2000, Curvelets: a surprisingly effective nonadaptive
representation for objects with edges: Curves and Surfaces Fitting, A Cohen, C Rabut,
L. Schumaker (eds.), Venderbilt University Press, Nashville, 105-120.
Candès, E.J. and Donoho, D., 2002, Recovering edges in ill-posed inverse problems:
Optimality of curvelet frames: Ann. Statist. 30 (2002), 784-842.
Candès, E.J. and Donoho, D. L., 2005, Continuous curvelet transform: I. resolution of the
wavefront set: Appl. Comput. Harmon. Anal.19-2,162-197.
Candes, E. J., J. K. Romberg, and T. Tao, 2006, Stable signal recovery from incomplete
and inaccurate measurements: Communications on Pure and Applied Mathematics, 59,
8, 1207-1223.
Cao, J., and Wu, R.S., 2005, Influence of Propagator and Acquisition Aperture on Image
Amplitude: Expanded abstracts, SEG 75th Annual Meeting, 1946-1949.
Cao, J. and Wu, R.S., 2008, Amplitude compensation of one-way wave propagators in
inhomogeneous media and its application to seismic imaging: Special issue
"Computational geophysics", Communications in computational physics, 3, No 1, 203-
221.
Cao, J. and Wu, R.S., 2009, Fast acquisition aperture correction in prestack depth
migration using beamlet decomposition, Geophysics, 74, S67-S74.
Červený, V., 1983, Synthetic body wave seismograms, for laterally varying structures by
the Guassian beam method, Geophys. J. R. astr. Soc., 73, 389-426.
Červený, V., 2001, Seismic ray theory: Cambridge University Press.
Červený, V., Popov, M.M. and Pšenčik, I., 1982, Computation of wave fields in
inhomogeneous media – Gaussian beam approach, Geophys. J. R. astr. Soc., 70, 109-
128.
Červený, V., L. Klimeš, and I. Pšenčik, 2007, Seismic ray method: Recent developments:
Advances in Geophysics, Vol 48, 48, 1-126.
Chauris, H., 2006, Seismic imaging in the curvelet domain and its implications for the
curvelet design: Expanded abstracts, SEG 76th Annual Meeting, 2404-2410.
Chen, L. and Wu, R.S., 2002, Target-oriented prestack beamlet migration using Gabor-
Daubechies frames: Expanded abstracts, SEG 72nd Annual Meeting, 1356-1359.
Chen, L., Wu, R.S. and Chen Y., 2006, Target-oriented beamlet migration based on
Gabor-Daubechies frame decomposition: Geophysics, 71, s37-s52.
Cohen, L., 1995, Time-frequency analysis: Prentice Hall PTR, New Jersey.
Cohen, A., 2003, Numerical Analysis of Wavelet Methods, North-Holland, Elsevier.
Coifman, R.R. and Meyer, Y., 1991, Remarques sur l’analyse de Fourier a fenetre,
Comptes Rendus de l’Academie des Sciences, Paris, Serie I 312, 259-261.
59
Coifman, R.R. and Wickerhauser, M.V., 1994, Wavelet and adapted waveform analysis,
in Wavelets, Mathematics and applications, edited by J.J. Benedetto and M.W. Frazier:
CRC Press, 399-423.
Coifman, R.R., Matviyenko, G. and Meyer, Y., 1997, Modulated Malvar-Wilson bases:
Appl. Comput. Harmon. Anal., 4, 58-61.
Chapman, C.H., 2004, Fundamentals of seismic wave propagation: Cambridge University
Press.
Chapman, C.H., and Drummond, R., 1982. Body-wave seismograms in inhomogeneous
media using Maslov asymptotic theory: Bull. seism. Soc. Am., 72, 277–317.
Choudhary, S. and Felsen, L.B., 1974, Analysis of Gaussian beam propagation and
diffraction of inyomog wave tracking: Bull Seis. Soc. Am., 72, 5277-5317.
Daubechies, I., 1988. Time-frequency localization operators: a geometric phase space
approach, IEEE Trans. Inform. Theory, 34, 605-612.
Daubechies, I., 1990, The wavelet transform, time-frequency localization and signal
analysis: IEEE Trans. Info. Theory, 36, 961-1005.
Daubechies, I., 1992, Ten Lectures on Wavelets: Philadelphia, Pennsylvania: Society for
Industrial and Applied Mathematics.
Daubechies, I. and Janssen, A.J.E.M., 1993, Two theorems on lattice expansions: IEEE
Trans. Information. Theory, 39(1), 3-6.
Daubechies, I., Jaffard, S. and Journé, J.-L., 1991, A simple Wilson orthonormal basis
with exponential decay: SIAM J. Math. Anal., 22, 554-573.
Demanet, L., 2006, Curvelets, wave atoms, and wave equations: Ph. D. Thesis, California
Inst. of Tech., California.
Demanet, L. and Ying, L., 2006, Wave atoms and sparsity of oscillatory patterns:
submitted.
Deschamps, G.A., 1971, Gaussian beams as a bundle of complex rays, Electronics Lett.,
7, 684-685.
De Hoop, M. and Stolk, C.C., 2002, Microlocal analysis of seismic inverse scattering in
anisotropic, elastic media: Comm. Pure Appl. Math. 55, 261-301.
Do, M. N, and Vertelli, M., 2003, Contourlet, in G. V. Welland, Beyound wavelets:
Academic Press.
Donoho, D., 1995, Nonlinear solution of linear inverse problems by wavelet-vaguelette
decomposition, Appl. Comput. Harmon. Analytic., 2, 101-126.
Donoho, D., 1999, Tight frames of k-plane ridgelets and the problem of representing
objects that are smooth away from d-dimensional singularities in n
: Proc. Nat. Acad.
Sci., USA, 96, 1828-1833.
Donoho, D. L., 2006, Compressed sensing: IEEE Transactions on Information Theory,
52, 4, 1289-1306.
Douma, H., and de Hoop, M.V., 2004, Wave-character preserving pre-stack map
migration using curvelets: Expanded Abstracts, 74th SEG Annual International
Meeting.
Douma, H., and de Hoop, M.V., 2005, On common-offset pre-stack time-migration with
curvelets: Expanded Abstracts, 75th SEG Annual International Meeting, 2009–2012.
Douma, H. and de Hoop, M.V., 2006, Leading-order seismic imaging using curvelets:
Expanded abstracts, SEG 76th Annual Meeting, 2411-2415.
60
Douma, H. and de Hoop, M.V., 2007, Leading-order seismic imaging using curvelets:
Geophysics 71, no. 1, S13-S28.
Einziger, P. D., S. Raz, and M. Shapira, (1986), Gabor Representation and Aperture
Theory: J. Opt. Soc. Am. A 3, 508–522.
Felsen, L.B., 1975, Complex rays, Philips Res. Rep., 30, 187-185.
Felsen, L.B., 1976, Complex-source-point solutions of the field equations and their
relation to the propagation and scattering of Guassian beams, Symposie Mathematics,
Inst. Nationa Alta Matematica, XVIII, 40-56, Academic Press.
Fishman, L. and McCoy, J.J., 1984, Derivation and application of extended parabolic
wave theories II. Path integral representations, J. Math. Phys., 25, 297-308.
Flatté, S.M., Dashen, R., Munk, W.H., Watson, K. and Zachariasen, F., 1979, Sound
transmission through a fluctuation ocean: Cambridge University Press.
Foster, D. and Huang, J., 1991, Global asymptotic solutions of the wave equation:
Geophys. J. Int., 105, 163-171.
Foster, D.J., Lane, F.D., Mosher, C.C. and Wu, R.S., 1997, Wavelet transforms for
seismic data processing: Expanded abstracts, SEG 67th Annual Meeting, 1318-1321.
Foster, D.J., Wu, R.S. and Mosher, C.C., 2002, Coherent-state solutions of the wave
equation, Expanded abstracts, SEG 72nd Annual Meeting, 1348-1351.
Freichtinger, H.G. and Strohmer, T., 1998, Gabor analysis and algorithms, Theory and
applications: Birkhäuser.
Geng, Y., R. S. Wu, and J. H. Gao, 2013, Gabor frame based Gaussian Packet migration,
Geophysical Prospecting, in press.
Herrmann, F., 2003, Optimal seismic imaging with curvelets: 70th Annual International
Meeting, SEG, Expanded Abstracts, 997–1000.
Herrmann, F. J., 2010, Randomized sampling and sparsity: Getting more information
from fewer samples: Geophysics, 75, 6, Wb173-Wb187.
Heyman, E. and B. Z. Steinberg, 1987, Spectral analysis of complex source pulsed
beams, J. Opt. Soc. Am. A 4, 473–480.
Gabor, D., 1946, Theory of Communication, J. Inst. Electr. Eng., London, 93(III), 429-
457.
Gao, J.H., Zhou Y., Mao, J., Chen, W., Wu R.S. and Li, Y., 2006, A wave propagation
method in local angle domain: Chinese Geophys. J., 50, 249-259.
Garding, L., 1987, Singularities in linear wave propagation: Lecture Notes in
Mathematics, 1241, Springer.
Geng, Y., R. S. Wu, and J. H. Gao, 2009, Dreamlet transform applied to seismic data
compression and its effects on migration: Expanded abstracts, SEG 79th Annual
Meeting, 28, 3640-3644.
Goupillaud, P., Grossmann, A., and Morlet, J., 1984, Cycle-octave and related transforms
in seismic signal analysis, Geoexploration 23, 85-102.
Grossman, J.P., 2005, Theory of adaptive, nonstationary filtering in the Gabor domain
with applications to seismic inversion: Ph.D. Thesis, University of Calgary, Canada.
Grossman, J. P., Margrave, G. F., Lamoureux, M. P., 2003, Adaptive, nonuniform Gabor
frames from partitions of unity: submitted to The Journal of Computational and
Applied Mathematics.
Grossmann, A., Morlet, J., 1984, Decomposition of Hardy functions into square
integrable wavelets of constant shape, SIAM J. Math. Anal., 15, pp. 723-736.
61
Grossman, A., Morlet, J., Paul, T., 1985, Transforms associated to square integrable
group representations I: general results, J. Math. Phys., 26, 2473-2479.
Grossman, A., Morlet, J., Paul, T., 1986, Transforms associated to square integrable
group representations II: examples, Ann. Inst. Henri Poincaré Physique théorique, 45,
293-309.
Grossman, A., Paul, T., 1984, Wave functions on subgroups of the group of affine
canonical transformations, in Resonances, Models and Phenomena, Lectures Notes in
Physics, Vol.211, Springer-Verlag, Berlin.
Haar, A., 1910, Zur Theorie der orthogonalen Functionensysteme, Math. Ann., 69, pp.
331-371.
Hale, D., 1992, Migration by the Kirchhoff, slant stack, and Gaussian beam methods:
Colorado School of Mines Center for Wave Phenomena Report 121.
Heil, C., Walnut, D., 1989, Continuous and discrete wavelet transforms, SIAM Rev., 31,
pp.628-666.
Hermanez-Figueroa, H., Zamboni-Rached, M., and Recami, R., 2008, “Localized
waves”, Wiley Series in Micorwave and Optical Engineering, Wiley-Interscience.
Herrmann, F., 2003, Optimal imaging with curvelets: Annual meeting of the SEG,
Expanded Abstracts.
Hill, N.R. 1990, Gaussian beam migration, Geophysics, 55, 1416-1428.
Hill, N.R. 2000, Prestack Gaussian beam depth migration, Geophysics, 66, 1240-1250.
Holschneider, M., 1995, Wavelets, an analysis tool: Clarendon Press, Chapter 5.
Jaffard, S., Meyer, Y., Ryan, R. D., 2001, Wavelets: Tool for Science and Technology:
Society for Industrial and Applied Mathematics, Philadelphia.
Jin, S., Luo, M.Q., Wu, R.S., and Walraven, D., 2005, Application of beamlet propagator
to migration amplitude correction: Expanded abstracts, SEG 75th Annual Meeting,
1962-1965.
Kaiser, G., 1994, A Friendly Guide to Wavelets: Birkhäuser, Boston, Basel, Berlin.
Kaiser, G., 2004, Eigenwavelets of the wave equation: Signals & waves, Austin, TX.
Keller, J.B. and Streifer, W., 1971, Complex rays with an application to Gaussian beams,
J. Opt. Soc. Am., 61, 40-43.
Kiselev, A. P., and M. V. Perel, 1999, Gaussian wave packets: Optics and Spectroscopy,
86, 3, 307-309.
Klauder, J.R., 1987, Global, uniform, asymptotic wave-equation solutions for large
wavenumbers: Annals of Physics, 180, 108-151.
Klauder, J.R. and Skagerstam, B.-S., 1985, Coherent States: World Scientific, Singapore.
Kleyn, A., 1977, On the migration of reflection time contour maps: Geophysical
Prospecting, 25, 125–140.
Klimeš, L., 1989, Gaussian packets in the computation of seismic wavefields:
Geophysical Journal International, 99, 2, 421-433.
Kravtsov, Yu. A., 1967, Complex rays and complex caustics, Radiophys. Quantum
Electronics, 10, 1283-1304.
Kravtsov, Yu. A. and Berczynski, P., 2007, Gaussian beams in inhomogeneous media: a
review: Stud. Geophys. Geod., 51, 1-36.
Li, S.X. and Liu, J.Q., 1994, Wavelet transform and the mathematical basis of inversion:
Geological Press of China (in Chinese).
62
Low, F., 1985, Complete sets of wave packets, in A Passion for Physics-Essays in Honor
of Geoffrey Chew, World Scientific, Singapore, pp. 17-22.
Luo, M. and Wu, R.S., 2003, 3D beamlet prestack depth migration using the local cosine
basis propagator, Expanded abstracts, SEG 73rd Annual Meeting, 985-988.
Luo, M., Wu, R.S., and Xie, X.B., 2004, Beamlet migration using local cosine basis with
shifting windows, Expanded abstracts, SEG 74th Annual Meeting, 945-948.
Luo, M.Q., Wu, R.S. and Xie, X.B., 2005, True amplitude one-way propagators
implemented with localized corrections on beamlets: Expanded abstracts, SEG 75th
Annual Meeting, 1966-1969.
Mao, J., and Wu, R.S., 2010, Target oriented 3D acquisition aperture correction in local
wavenumber domain, 80th Annual International Meeting, SEG, Expanded Abstracts
29, 3237-3241.
Mao, J. and Wu, R.S., 2013, Target oriented acquisition aperture correction for 3D
subsurface imaging with beamlet migration, Geophysical Prospecting, in press.
Margrave, G. F. and Ferguson, R.J., 1999, Wavefield extrapolation by nonstationary
phase shift: Geophysics, 64, 1067-1078.
Malvar, H.S., 1990, Lapped transforms for efficient transform/subband coding: IEEE
Trans. Acoust. Speech Signal Process., 38, 969-978.
Malvar, H.S., 1992, Signal processing with lapped transforms, Artech House, Norwood,
MA.
Mallat, S., 1989, A theory for multiresolution signal decomposition: The wavelet
representation, IEEE Trans. Patt. Analytic. Mach. Intell., 11, 674-693.
Mallat, S., 1999, A Wavelet Tour of Signal Processing, second edition: Academic Press.
Mallat, S. and Zhang, Z., 1993, Matching pursuits with time-frequency dictionaries, IEEE
Trans. Signal Process., 41, 3397-3415.
Meyer, F. G., 1998, Image compression in libraries of bases, [Lecture notes for a course
given at the Institut Henri Poincaré, Paris]
Meyer, Y., Ondelettes et Opérateurs I: Ondelettes, Hermann, Paris, 1990(in French).
Wavelets and Operators, Cambridge Univ. Press, Cambridge, U. K., 1992(in English).
Meyer, Y. and Coifman, R.R., 1991, Ondelettes et Opérateurs III: Opérateurs
multilinéaires, Hermann, Paris, (in French). Wavelets, Cambridge Univ. Press,
Cambridge, U. K., 1997(in English).
Meyer, Y. and Coifman, R.R., 1997, Brushlets : A tool for directional image analysis and
image compression : Appl. Comput. Harmon. Anal., 4, 147-187.
Meyer, Y. and Coifman, R.R., 1997, Wavelets, Calderόn-Zygmund and Multilinear
Operators: Cambridge University Press, Cambridge.
Morlet J., Arens, G., Fourgeau, S, E. and Giard, D., 1982a, Wave propagation and
sampling theory -Part I: Complex signal and scattering in multilayered media:
Geophysics, 47, 203-221.
Morlet J., Arens, G., Fourgeau, S, E. and Giard, D., 1982b, Wave propagation and
sampling theory -Part II: Sampling theory and complex waves : Geophysics, 47, 222-
236.
Mosher, C.C, Foster, D.J. and R.S. Wu, 1996, Phase shift migration with wave packet
algorithms, “Mathematical methods in geophysical imaging IV”, Proc. SPIE, 2822, 2-
16.
63
Nowack, R.L., 2003, Calculation of synthetic seismograms with Gaussian beams: Pure
Appl. Geophys. 160, 487-507.
Nowack, R.L. S. Dasgupta, G.T. Schuster and J.M. Sheng, 2006, Correlation migration
using Gaussian beams of scattered teleseismic body waves: Bull. Seism. Soc. Am.,
Vol. 96, pp. 1-10.
Norris, A. N., B. S. White, and J. R. Schrieffer, 1987, Gaussian wave-packets in
inhomogeneous-media with curved interfaces: Proceedings of the Royal Society of
London Series a-Mathematical Physical and Engineering Sciences, 412, 1842, 93-123.
Pascal, A., Guido, W. and Wickerhauser M.V., 1992, Local sine and cosine bases of
Coifman and Meyer and the construction of smooth wavelets, Wavelets: A Tutorial in
Theory and Applications (ed. Charles K. Chui), 237-256, Academic Press, Inc.
Perel, M. V., and M. S. Sidorenko, 2007, New physical wavelet 'Gaussian wave packet':
Journal of Physics a-Mathematical and Theoretical, 40, 13, 3441-3461.
Popov, M.M., 1982, A new method of computation of wave fields using Gaussian beams:
Wave motion, 4, 85-97.
Popov, M.M., 2002, Ray theory and Gaussian beam method for geophysicists: EDUFBA,
Salvador-Bahia.
Qian, J. L., and L. X. Ying, 2010, Fast gaussian wavepacket transforms and gaussian
beams for the schrodinger equation: Journal of Computational Physics, 229, 20, 7848-
7873.
Raslton, J., 1983, Gaussian beams and the propagation of sigularities: Littman W.(ed):
Studies in Partial Differential Equations, MAA Studies in Mathematics, 23, 206-248.
Raz, S., 1987, Beam stacking: a generalized preprocessing technique: Geophysics 52,
1199-1210.
Ren, H., Wu, R.S. and Wang, H., 2011, Wave equation least square imaging using the
local angular Hessian for amplitude correction, Geophysical Prospecting, , 59, 651-
661.
Ristow, D., and Rühl, T. (1994), Fourier finite-difference migration, Geophysics 59,
1882-1893.
Sava, P. and Fomel, S., 2002, Angle-domain common-image gathers by wavefield
continuation methods: ???
Smith, H., 1998a, A Hardy space for Fourier integral operators. J. Geom. Anal. 8, 629-
653.
Smith, H., 1998b, A parametrix construction for wave equations with coefficients. Ann.
Inst. Fourier (Grenoble) 48, 797-835.
Stein, E. M., 1993, Harmonic analysis: real-variable methods, orthogonality, and
oscillatory integrals: Princeton University Press.
Steinberg, B.Z., Heyman, E. and Felsen, L.B., 1991, Phase-space beam summation for
time-dependent radiation from large apertures: continuous parameterization: J. Opt.
Soc. Am. 8, 943-958.
Steinberg, B.Z. and Heyman, E., 1991, Phase-space beam summation for time-dependent
radiation from large apertures: discretized parameterization: J. Opt. Soc. Am. 8, 959-
966.
Steinberg, B.Z., 1993, Evolution of local spectra in smoothly varying nonhomogeneous
environments-Local canonization and marching algorithms, J. Acoust. Soc. Am. 93,
2566-2580.
64
Steinberg, B.Z. and McCoy, J.J., 1993, Marching acoustic fields in a phase space: J.
Acoust. Soc. Am. 93, 188-204.
Steinberg, B.Z. and R. Birman, 1995, Phase-space marching algorithm in the presence of
a planar wave velocity discontinuity-A Qualitative study, J. Acoust. Soc. Am. 98, 484-
494.
Strömberg, J.-O., 1983, A modified Franklin system and higher-order spline systems on
as unconditional bases for Hardy spaces, in conference on Harmonic Analysis in
Honor of Antoni Zygmund, vol. II, W. Beckner et al., eds., Wadsworth, Belmont, CA,
pp. 475-494.
Sweldens, W. and Shröder, P., 1996, Building Your Own Wavelets at Home.
Thomson, C.J., 2001, Seismic coherent states and ray geometric spreading: Geophys. J.
Int., 144, 320-342.
Thomson, C.J., 2004, Coherent states analysis of the head wave problem: an
overcomplete representation and its relationship to rays and beams: Geophys. J. Int.,
157, 1189-1205.
Torrésani, B., 1991, Wavelets associated with representations of the affine Weyl-
Heisenberg group, J. Math. Phys. 32, 1273-1279.
Van den Berg, 2004, Wavelets in Physics : Cambridge University Press.
Ville, J., 1948, Théorie et applications de la notion de signal analytique, Câbles et
Transmissions, Laboratoire de Télécommunications de la Société Alsacienne de
Construction Mécanique, 2A, pp. 61-74.
Wang, Y. and Wu, R.S., 1998a, Migration operator decomposition and compression
using a new wavelet packet best basis algorithm, Expanded abstracts, SEG 68th
Annual Meeting, 1167-1170.
Wang, Y. and Wu, R.S., 1998b, Decomposition and compression of Kirchhoff migration
operator by adapted wavelet packet transform, Wavelet Applications in Signal and
Image Processing VI, Proc., SPIE, 3458, 246-258.
Wang, Y. and Wu, R.S., 2002, Beamlet prestack depth migration using local cosine basis
propagator, Expanded abstracts, SEG 72nd Annual Meeting, 1340-1343.
Wang, Y., Cook, R., and Wu, R.S., 2003, 3D local cosine beamlet propagator, Expanded
abstracts, SEG 73rd Annual Meeting, 981-984.
Wang, Y., Verm, R., and Bednar, B., 2005, Application of beamlet migration to the
SmaartJV Sigsbee2A model, Expanded abstracts, SEG 75th Annual Meeting, 1958-
1961.
Weber, M., 1955, Die bestimmung einer beliebig gekruemmten schichtgrenze aus
seismischen reflexionsmessungen: Geofisica Pura e Applicata, 32, 7–11.
Wickerhauser, M.V., 1993, Smooth localized orthonormal bases: Comptes Rendus de
l’Academie des Sciences de Paris, 316, 423-427.
Wickerhauser, M.V., 1994, Adapted wavelet analysis form theory to software: A K
Peters.
Wigner, E.P., 1932, On the quantum correction for thermodynamic equilibrium, Phys.
Rev., 40, pp. 749-759.
Wilson, K.G., 1971, Renormalization group and critical phenomena II; Phase-space cell
analysis of critical behavior, Phys. Rev. B. 4, pp. 3184-3205.
65
Wu, B., R.S. Wu, and J. Gao, 2009, Dreamlet prestack depth migration using local cosine
basis and local exponential frames, Expanded Abstracts, SEG 79th Annual Meeting,
2753-2757.
Wu, B., Wu, R.-S. & Gao, J.-H., 2013, Dreamlet source-receiver survey sinking prestack
depth migration, Geophysical Prospecting, 61, 63-64.
Wu, R. S., 1985, Gaussian beams, complex rays, and the analytic extension of the
Green’s function in smoothly inhomogeneous media, Geophys. J. R. astr. Soc., 83, 93-
110.
Wu, R. S., and Aki, K, 1988, Seismic wave scattering in the three-dimensionally
heterogeneous earth, in the special issue "Seismic Wave Scattering and Attenuation",
edited by Wu and Aki, Pure and Applied Geophys., 128, 1-6.
Wu, R. S., and Chen, L, 2001, Beamlet migration using Gabor-Daubechies frame
propagator: 63rd Conference & Technical Exhibition, EAGE, Expanded abstracts,
P.74.
Wu, R. S., and Chen, L., 2002a, Wave propagation and imaging using Gabor-Daubechies
beamlets: Theoretical and Computational Acoustics, World Scientific, New Jersey,
661-670.
Wu, R. S. and Chen, L., 2002b, Mapping directional illumination and acquisition-
aperture efficacy by beamlet propagators: 72nd Ann. Internat. Mtg., Soc. Expl.
Geophys., Expanded Abstracts, 1352-1355.
Wu, R.S. and Chen, L., 2003, Directional illumination and acquisition dip-response:
Extended Abstracts, EAGE 65rd Annual Meeting.
Wu, R. S. and Chen, L., 2006, Directional illumination analysis using beamlet
decomposition and propagation: Geophysics, 71, s147-s159.
Wu, R. S., Chen, L., and Wang, Y., 2002, Prestack migration/imaging using synthetic
beamsources and plane sources, Stud. Geophys. Geod., 46, 651-665.
Wu, R.S., Chen, S. and Luo, M., 2004, Migration amplitude correction in angle domain
using beamlet decomposition: Expanded abstracts, EAGE 66th Annual Meeting, G029.
Wu, R.S., Geng, Y. and Wu, B., 2011, Physical wavelet defined on an observation plane
and the Dreamlet, 81st Annual International Meeting, SEG, Expanded Abstracts, 3835-
3839.
Wu, R.S. and Jin, S., 1997, Windowed GSP (generalized screen propagators) migration
applied to SEG-EAEG salt model data, Expanded abstracts, SEG 67th Annual
Meeting, 1746-1749.
Wu, R. S., and Jin, S., 1997, Windowed GSP (generalized screen propagators) migration
applied to SEG-EAEG salt model data: 67th Ann. Internat. Mtg., Soc. Expl. Geophys.,
Expanded Abstracts, 1746-1749.
Wu, R.S., Luo, M., Chen, S. and Xie, X.B., 2004, Acquisition aperture correction in
angle-domain and true-amplitude imaging for wave equation migration: Expanded
abstracts, SEG 74th Annual Meeting, 937-940.
Wu, R.S., and Luo, M.Q., 2005, Comparison of different scheme of image amplitude
correction in prestack depth migration: Expanded abstracts, SEG 75th Annual
Meeting, 2060-2063.
Wu, R.S. and Maupin, V., 2007, Advances in Wave Propagation in Heterogeneous Earth:
Vol. 48 of “Advances in Geophysics”, Series Editor, Dmowska, R.: Elsevier.
66
Wu, R.S. and Wang, Y., 1998, Comparison of propagator decomposition in seismic
imaging by wavelets, wavelet-packets, and local harmonics, Mathematical Methods in
Geophysical Imaging, V, Proc. SPIE, 3453, 163-179.
Wu, R. S., Wang, Y. and Gao, J. H., 2000, Beamlet migration based on local perturbation
theory, 70th Ann. Internat. Mtg., Soc. Expl. Geophys., Expanded abstracts, 1008-1011.
Wu, R.S., Wang, Y. and Luo, M., 2003, Local-cosine beamlet migration for 3D complex
structures, Eighth International Congress of the Brazilian Geophysical Society, 14-18.
Wu, R.S., Wang, Y. and Luo, M., 2008, Beamlet migration using local cosine basis,
Geophysics, 73, S207-217.
Wu, R.S., Wu, B.Y. and Geng, Y. 2008, Seismic wave propagation and imaging using
time-space wavelets, Expanded abstracts, SEG 78th Annual Meeting, 2983-2987.
Wu, R.S., B. Wu, and Y. Geng, 2009, Imaging in compressed domain using
dreamlets, CPS/SEG Beijing ’2009, International Geophysical Conference, Expanded
Abstracts, ID: 57.
Wu, R.S., Xie, X.B. and Jin, S., 2012, One-Return Propagators and The Applications in
Modeling and Imaging, Chapter 2 in “Imaging, Modeling and Assimilation in
Seismology”, Higher Education Press Limited Company, Beijing, 65-105.
Wu, R.S. and F. Yang, 1997, Seismic imaging in wavelet domain: Decomposition and
compression of imaging operator, "Wavelet Applications in Signal and Image
Processing V", Proc. SPIE, 3169, 148-162.
Wu, R.S., F. Yang, Z. Wang and L. Zhang, 1997, Migration Operator Compression by
Wavelet Transform: Beamlet Migrator, Expanded Abstracts of the Technical Program,
SEG 67th Annual Meeting, 1646-1649.
Xie, X.B. and Wu, R.S., 2002, Extracting angle related image from migrated wavefield:
Expanded abstracts, SEG 72nd Annual Meeting, 1360-1363.
Xie X.B. and Wu, R.S., 2003, Three-dimensional illumination analysis using wave
eauation based propagator: Expanded abstracts, SEG 73rd Annual Meeting, 989-992.
Xie, X.B., Jin, S. and Wu, R.S., 2006, Wave-equation based seismic illumination
analysis, Geophysics, 71, No 5, S169-177.
Xie, X.B., Jin, S. and Wu, R.S., 2004, Wave equation based illumination analysis:
Expanded abstracts, SEG 74th Annual Meeting, 933-936.
Xu, S. and Lambare, G., 1998, Maslov + Born migration/inversion in complex media,
Expanded abstracts, SEG 68th Annual Meeting, 1702-1707.
Ying, L., Demanet, L., Candès, E.J., 3D Curvelet Transform: Proc. Conf. Wavelets XI,
San Diego, 2006.
Young, R.K., 1993, Wavelet theory and its applications: Kluwer Academic Publishers.
Zacek, K., 2004, Gaussian packet pre-stack depth migration: Expanded Abstracts: 74th
Annual International Meeting, SEG, 957-960.
Zacek, K., 2005, Gaussian packet pre-stack depth migration of the marmousi data set:
Expanded Abstracts: 75th Annual International Meeting, SEG, 1822-1925.
Zacek, K., 2006a, Decomposition of the wave field into optimized gaussian packets:
Studia Geophysica Et Geodaetica, 50, 3, 367-380.
Zacek, K., 2006b, Optimization of the shape of gaussian beams: Studia Geophysica Et
Geodaetica, 50, 3, 349-366.
67
Author Information
Ru-Shan Wu
Modeling and Imaging Laboratory, Institute of Geophysics and Planetary
Physics/Department of Earth and Planetary Sciences, University of California, Santa
Cruz, CA 95064, USA
E-mail: rwu@ucsc.edu
Jinghuai Gao
School of Electronic and Information Engineering, Xi’an Jiaotong University, Xi’an,
China