Carnegie Mellon Universitycschafer/cmspbs.pdf · 2009. 2. 26. · Carnegie Mellon University
The Frequency Domain - Carnegie Mellon University
Transcript of The Frequency Domain - Carnegie Mellon University
![Page 1: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/1.jpg)
The Frequency Domain
15-463: Computational Photography
Alexei Efros, CMU, Spring 2010
Somewhere in Cinque Terre, May 2005
Many slides borrowed
from Steve Seitz
![Page 2: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/2.jpg)
Salvador Dali
“Gala Contemplating the Mediterranean Sea,
which at 30 meters becomes the portrait
of Abraham Lincoln”, 1976
![Page 3: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/3.jpg)
![Page 4: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/4.jpg)
![Page 5: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/5.jpg)
A nice set of basis
This change of basis has a special name…
Teases away fast vs. slow changes in the image.
![Page 6: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/6.jpg)
Jean Baptiste Joseph Fourier (1768-1830)
had crazy idea (1807):
Any periodic function
can be rewritten as a
weighted sum of sines
and cosines of different
frequencies.
Don’t believe it? • Neither did Lagrange,
Laplace, Poisson and
other big wigs
• Not translated into
English until 1878!
But it’s true!• called Fourier Series
![Page 7: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/7.jpg)
A sum of sines
Our building block:
Add enough of them to get
any signal f(x) you want!
How many degrees of
freedom?
What does each control?
Which one encodes the
coarse vs. fine structure of
the signal?
xAsin(
![Page 8: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/8.jpg)
Fourier Transform
We want to understand the frequency of our signal. So,
let’s reparametrize the signal by instead of x:
xAsin(
f(x) F( )Fourier
Transform
F( ) f(x)Inverse Fourier
Transform
For every from 0 to inf, F( ) holds the amplitude A
and phase of the corresponding sine • How can F hold both? Complex number trick!
)()()( iIRF22 )()( IRA
)(
)(tan 1
R
I
We can always go back:
![Page 9: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/9.jpg)
Time and Frequency
example : g(t) = sin(2pf t) + (1/3)sin(2p(3f) t)
![Page 10: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/10.jpg)
Time and Frequency
example : g(t) = sin(2pf t) + (1/3)sin(2p(3f) t)
= +
![Page 11: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/11.jpg)
Frequency Spectra
example : g(t) = sin(2pf t) + (1/3)sin(2p(3f) t)
= +
![Page 12: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/12.jpg)
Frequency Spectra
Usually, frequency is more interesting than the phase
![Page 13: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/13.jpg)
= +
=
Frequency Spectra
![Page 14: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/14.jpg)
= +
=
Frequency Spectra
![Page 15: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/15.jpg)
= +
=
Frequency Spectra
![Page 16: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/16.jpg)
= +
=
Frequency Spectra
![Page 17: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/17.jpg)
= +
=
Frequency Spectra
![Page 18: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/18.jpg)
= 1
1sin(2 )
k
A ktk
Frequency Spectra
![Page 19: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/19.jpg)
Frequency Spectra
![Page 20: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/20.jpg)
Extension to 2D
in Matlab, check out: imagesc(log(abs(fftshift(fft2(im)))));
![Page 21: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/21.jpg)
Man-made Scene
![Page 22: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/22.jpg)
Can change spectrum, then reconstruct
![Page 23: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/23.jpg)
Low and High Pass filtering
![Page 24: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/24.jpg)
The Convolution Theorem
The greatest thing since sliced (banana) bread!
• The Fourier transform of the convolution of two functions is the product of their Fourier transforms
• The inverse Fourier transform of the product of two Fourier transforms is the convolution of the two inverse Fourier transforms
• Convolution in spatial domain is equivalent to multiplication in frequency domain!
]F[]F[]F[ hghg
][F][F][F 111 hggh
![Page 25: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/25.jpg)
2D convolution theorem example
*
f(x,y)
h(x,y)
g(x,y)
|F(sx,sy)|
|H(sx,sy)|
|G(sx,sy)|
![Page 26: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/26.jpg)
Fourier Transform pairs
![Page 27: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/27.jpg)
Low-pass, Band-pass, High-pass filters
low-pass:
High-pass / band-pass:
![Page 28: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/28.jpg)
Edges in images
![Page 29: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/29.jpg)
What does blurring take away?
original
![Page 30: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/30.jpg)
What does blurring take away?
smoothed (5x5 Gaussian)
![Page 31: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/31.jpg)
High-Pass filter
smoothed – original
![Page 32: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/32.jpg)
Band-pass filtering
Laplacian Pyramid (subband images)Created from Gaussian pyramid by subtraction
Gaussian Pyramid (low-pass images)
![Page 33: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/33.jpg)
Laplacian Pyramid
How can we reconstruct (collapse) this
pyramid into the original image?
Need this!
Original
image
![Page 34: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/34.jpg)
Why Laplacian?
Laplacian of Gaussian
Gaussian
delta function
![Page 35: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/35.jpg)
Unsharp Masking
200 400 600 800
100
200
300
400
500
- =
=+
![Page 36: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/36.jpg)
Image gradient
The gradient of an image:
The gradient points in the direction of most rapid change in intensity
The gradient direction is given by:
• how does this relate to the direction of the edge?
The edge strength is given by the gradient magnitude
![Page 37: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/37.jpg)
Effects of noise
Consider a single row or column of the image• Plotting intensity as a function of position gives a signal
Where is the edge?
How to compute a derivative?
![Page 38: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/38.jpg)
Where is the edge?
Solution: smooth first
Look for peaks in
![Page 39: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/39.jpg)
Derivative theorem of convolution
This saves us one operation:
![Page 40: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/40.jpg)
Laplacian of Gaussian
Consider
Laplacian of Gaussian
operator
Where is the edge? Zero-crossings of bottom graph
![Page 41: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/41.jpg)
2D edge detection filters
is the Laplacian operator:
Laplacian of Gaussian
Gaussian derivative of Gaussian
![Page 42: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/42.jpg)
Try this in MATLAB
g = fspecial('gaussian',15,2);
imagesc(g); colormap(gray);
surfl(g)
gclown = conv2(clown,g,'same');
imagesc(conv2(clown,[-1 1],'same'));
imagesc(conv2(gclown,[-1 1],'same'));
dx = conv2(g,[-1 1],'same');
imagesc(conv2(clown,dx,'same'));
lg = fspecial('log',15,2);
lclown = conv2(clown,lg,'same');
imagesc(lclown)
imagesc(clown + .2*lclown)
![Page 43: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/43.jpg)
Campbell-Robson contrast sensitivity curve
![Page 44: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/44.jpg)
Depends on Color
R G B
![Page 45: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/45.jpg)
Lossy Image Compression (JPEG)
Block-based Discrete Cosine Transform (DCT)
![Page 46: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/46.jpg)
Using DCT in JPEG
The first coefficient B(0,0) is the DC component,
the average intensity
The top-left coeffs represent low frequencies,
the bottom right – high frequencies
![Page 47: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/47.jpg)
Image compression using DCT
DCT enables image compression by
concentrating most image information in the
low frequencies
Loose unimportant image info (high
frequencies) by cutting B(u,v) at bottom right
The decoder computes the inverse DCT – IDCT
•Quantization Table3 5 7 9 11 13 15 17
5 7 9 11 13 15 17 19
7 9 11 13 15 17 19 21
9 11 13 15 17 19 21 23
11 13 15 17 19 21 23 25
13 15 17 19 21 23 25 27
15 17 19 21 23 25 27 29
17 19 21 23 25 27 29 31
![Page 48: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/48.jpg)
Block size in JPEG
Block size
• small block
– faster
– correlation exists between neighboring pixels
• large block
– better compression in smooth regions
• It’s 8x8 in standard JPEG
![Page 49: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/49.jpg)
JPEG compression comparison
89k 12k
![Page 50: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/50.jpg)
Morphological Operation
What if your images are binary masks?
Binary image processing is a well-studied field,
based on set theory, called Mathematical
Morphology
![Page 51: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/51.jpg)
Preliminaries
![Page 52: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/52.jpg)
Preliminaries
![Page 53: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/53.jpg)
Preliminaries
![Page 54: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/54.jpg)
Basic Concepts in Set Theory
A is a set in , a=(a1,a2) an element of A, a A
If not, then a A
: null (empty) set
Typical set specification: C={w|w=-d, for d D}
A subset of B: A B
Union of A and B: C=A B
Intersection of A and B: D=A B
Disjoint sets: A B=
Complement of A:
Difference of A and B: A-B={w|w A, w B}=
Z 2
Ac {w |w A}
A Bc
![Page 55: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/55.jpg)
Preliminaries
} ,|{)(
} ,|{ˆ
AaforzaccA
BbforbwwB
z
![Page 56: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/56.jpg)
Dilation and Erosion
Two basic operations:• A is the image, B is the “structural element”, a mask akin to a kernel
in convolution
Dilation :
(all shifts of B that have a non-empty overlap with A)
Erosion :
(all shifts of B that are fully contained within A)
}])[(|{
})(|{
AABzBA
ABzBA
z
z
})(|{ ABzBA z
![Page 57: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/57.jpg)
Dilation
![Page 58: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/58.jpg)
Dilation
![Page 59: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/59.jpg)
Erosion
![Page 60: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/60.jpg)
Erosion
Original image Eroded image
![Page 61: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/61.jpg)
Erosion
Eroded once Eroded twice
![Page 62: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/62.jpg)
Opening and Closing
Opening : smoothes the contour of an object, breaks narrow
isthmuses, and eliminates thin protrusions
Closing : smooth sections of contours but, as opposed to opning, it
generally fuses narrow breaks and long thin gulfs, eliminates
small holes, and fills gaps in the contour
Prove to yourself that they are not the same thing. Play around with bwmorph in Matlab.
BBABA )(
BBABA )(
![Page 63: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/63.jpg)
OPENING: The
original image
eroded twice and
dilated twice
(opened). Most
noise is removed
Opening and Closing
CLOSING: The
original image
dilated and then
eroded. Most
holes are filled.
![Page 64: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/64.jpg)
Opening and Closing
![Page 65: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/65.jpg)
Boundary Extraction
)()( BAAA
![Page 66: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/66.jpg)
Boundary Extraction
![Page 67: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/67.jpg)
Project #2: Miniatures!
![Page 68: The Frequency Domain - Carnegie Mellon University](https://reader035.fdocuments.us/reader035/viewer/2022062310/62a77e03b225c15c53141425/html5/thumbnails/68.jpg)
Project #2: Fake Miniatures!