Evaluation of Spectral Issues in Sparse Aperture Imaging ... · TerraPix Digital Aerial Camera 4080...
Transcript of Evaluation of Spectral Issues in Sparse Aperture Imaging ... · TerraPix Digital Aerial Camera 4080...
R I T Rochester Institute of Technology
Evaluation of Spectral Issues inEvaluation of Spectral Issues inSparse Aperture Imaging SystemsSparse Aperture Imaging Systems
R.E. Introne
Digital Imaging and Remote Sensing LaboratoryChester F. Carlson Center for Imaging Science
Rochester Institute of Technology
The views expressed in this briefing are those of the author and do not reflect the policy or position of theUnited States Air Force, Department of Defense, or the U.S. Government.
R I T Rochester Institute of Technology
OverviewOverview
• Theory
• Approach
• Results
• Summary
R I T Rochester Institute of Technology
Sparse Aperture SystemsSparse Aperture Systems
Goal: Synthesize a larger effective collection aperture with a configuration of smaller phased subapertures
Multiple Phased Telescopes
Secondary
Primary
Tertiary
Focal Plane
Common Secondary Mirror
•• Includes multiple aperture & common secondary,Includes multiple aperture & common secondary,sparse & dilute telescope conceptssparse & dilute telescope concepts
•• Emphasize extended scene, remote sensing applications Emphasize extended scene, remote sensing applications
Image courtesy of HarveyImage courtesy of Boucher
Image courtesy of Fiete
R I T Rochester Institute of Technology
Research ObjectivesResearch Objectives
• Develop understanding of Sparse Aperture phenomenology– System-level diffraction (OTF/PSF) effects
– Photon, dark current & read noise effects
– Image restoration techniques
• Develop theoretical basis for modeling spectrally diverse SparseAperture system performance
– Target source spectral radiance
– Polychromatic system OTF
– Spectrally diverse noise
• Implement strawman model for spectrally diverse Sparse Aperture collection systems
– Scene spectral radiance
– Polychromatic system OTF
– Spectrally diverse system noise
R I T Rochester Institute of Technology
Signal ModelSignal Model
[ ] [ ] [ ] [ ]λλλλ ,,,,,,,, objincoh yxnyxyxfyxg +∗= h
Imaging System (Frequency Domain)
Imaging System (Space Domain)
where:
[ ] [ ] [ ] [ ]ληξληξληξληξ ,,,,,,,, objincoh NFG +⋅= H
Fourier Transform Pairs
[ ] [ ]{ }λληξ ,,,OTF x,yhFH =≡[ ]λ,,PSF yxh≡
[ ] [ ] [ ]( )
[ ] [ ] ( ) ( ) λλληλτληξληξπ
ληξληξηξ dLhcfFTA
FS fillopt
0FTsource,2
intdetobj ,,,,
#4,,,,,out
freq ∫∞
=⋅= HH
[ ] [ ] [ ] [ ] ( ) [ ] [ ]yxnyxnTyxnyxyxfyxn ,,,,,,,,, 3read2intdc1obj σσλλλ ++∗≅ h
Noise Model
Signal Frequency Spectrum
incohADC
elecconvcounts
2 gSGGI
n
=
Quantization
R I T Rochester Institute of Technology
Image RestorationImage Restoration
1.0
ρco
[ ]ηξ ,MTF H=
1.0
ρco
threshold
Inverse Filter
Pseudo-Inverse Filter
ξ ξ
Transfer Function Inverse Filtering Wiener Filtering
1.0
ρco
Inverse Filter
Wiener Filter
ξ
[ ] [ ] [ ]
[ ]
<
≥=
threshold, when 0
threshold, when ,
1,pseudo
ηξ
ηξηξηξ
H
HHW
[ ] [ ]ηξηξ
,1,inverse H
=W
• Inverse Filters tend to over-boost the noise or be non-optimum at high frequencies
• Optimum solution from a mean-square error viewpoint is the Wiener-Helstrom Filter
[ ] [ ][ ] [ ]
[ ]ηξηξηξ
ηξηξ
,,,
,,
f
n2
*wiener
SSW
+=
H
H
R I T Rochester Institute of Technology
Modeling Approach OverviewModeling Approach Overview
PupilFunction
PhaseErrors
UnaberratedOptics OTF
OpticsOTF
DetectorMTF
ObjectRadiance
DetectorNoise
RestorationFilter
RestoredImage
X
MotionMTF
X XX +
DIRSIG
MODTRAN
X
AberratedPSF
UnaberratedPSF
{}⋅F {}⋅-1F RawImage {}⋅-1F
{}⋅-1F
{}⋅-1F
{}⋅F
{}2⋅F
{}2⋅F
{}∑ ⋅λ
iλ iλ iλ iλ iλ iλ iλ
WavefrontError
Estimate
ObjectEstimate
R I T Rochester Institute of Technology
Strawman Collection ScenarioStrawman Collection Scenario
Panchromatic(polychromatic world)
Sparse Aperture Configurations:Annulus, Tri-arm, Golay-6
Pupil Function Phase Profile
Scenario #1Parameter Value
Spectral passband 0.4-0.7 µmSystem f-number (f#) 18.0Optical sampling (Q=λf/pD) 2.0 (Nyquist)Ground Sample Distance (GSD) 0.4572 m (18 in)Optical Transmission (τopt) 0.9Secondary Obscuration (εsub) 0.24Focal Plane Array (FPA) Staring Frame CCDRead Noise 50 rms electronsDynamic Range 11 bitsImage Smear 0.25 pixelsrms Wavefront Error 0.12 waves (λ = 0.65 µm)Atmosphere MODTRAN 4.0 mid-latitude summerVisibility 17.0 kmSimulation Time ~1700 GMT (~79° sun angle)Target Location 43.2° N Lat, 77.6° W Lon
R I T Rochester Institute of Technology
Orbital AnalysisOrbital Analysis
Orbit design and collection geometry evaluation performed
through use of STK
1 Jun 2003 16:48:00.00
Access Summary Report
Start Time (UTCG) Stop Time (UTCG)---------------------- ----------------------1 Jun 2003 16:33:33.66 1 Jun 2003 17:41:17.55
Duration (sec)--------------4063.893
R I T Rochester Institute of Technology
Imagery Collection GeometryImagery Collection Geometry
STK enables detailed evaluation of vehicle pointing (az, el) and position
(lat, lon, alt) required for DIRSIG
Performed arbitrary simulated nadir ascending pass at
~1300 local time on 1 Jun 03
R I T Rochester Institute of Technology
Suburban Scene RadiometrySuburban Scene RadiometryRochester Megascene
– Complex scene representative of typical remotely sensed targets
– Greater Rochester suburban area modeled with good fidelity
– Advertised to be accurate at 1-m (39.4-in) Ground Sample Distance
– Underlying texture map sampled at 6-in GSD
– Features within scene modeled well below 6 inches
Scene Attributes:
Simulations were performed using collection geometry acquired from STK and oversampled by 3:1(6 in to support 18-in GSD in sparse aperture simulation)
RGB spectral radiance image: 0.65µm (R), 0.55µm (G), 0.45µm (B)
1 Jun 2003 12:56:59.00
R I T Rochester Institute of Technology
Rural Scene RadiometryRural Scene RadiometryRochester Microscene
1 Jun 2003 12:56:59.00
Simulations were performed using collection geometry acquired from STK and oversampled between 3:1(6 in) and 18:1 (1 in)
Concealed MilitaryVehicles
R I T Rochester Institute of Technology
Alternative Scene OptionsAlternative Scene Options
TerraPix Digital Aerial Camera
4080 x 4080 pixel frame
0.4 – 0.9 µmRGB
~7 in GSD
12 bits
CitiPix Aerocolor Film Camera
9.5in film
0.38 – 0.7 µmRGB
~6 in GSD
8 bits (digitized)
Hymap Hyperspectral Sensor
512 pixel scanner
0.45 – 2.5 µm144 hyperspectral bands
~20 ft GSD
16 bits
Potential Aerial Object Scenes
R I T Rochester Institute of Technology
Complex Exit PupilComplex Exit Pupil
[ ] [ ]
= yxwiyxpyx ,2exp,],[λπ
p
[ ] [ ]
−−−−= ∑
=iii
N
iiii yyxxwiyyxxpyx ,2exp,],[
1 λπ
p
Conventional Pupil Function :
Sparse Aperture Pupil Function :
[ ] [ ] ( )
( ) ( )
( )
sorder termhigher
,,,
406
305
22204
20
23
2202
2221
20302
2210
distortion curvature field
mastigmatis coma spherical
piston tilt defocus
+
++++
+++++
+++=≡
xbxxbyxxb
xxbyxxxbyxb
xaxxayxaxyxwyxw
Aberration (wavefront error) function:
R I T Rochester Institute of Technology
Imaging System OTFImaging System OTF
[ ]∏=
=N
ii
1sys ,,MTFMTF ληξ
[ ] [ ] [ ] [ ] [ ] [ ]ληξηξηξηξληξληξ ,,MTF,MTF,MTF,MTF,,OTF,,OTF diffjittersmeardetapsys ⋅⋅⋅⋅=
[ ] [ ] [ ]
[ ]dxdyyxp
zzzz
∫ ∫∞+
∞−
=
,
, ,,,OTF 2222ap
ηλξληλξλληξ pp 9
System Transfer Function developed by cascading individual component OTF :
[ ]∏=
=N
ii
1sys ,,OTFOTF ληξ
Optical Transfer Function Modulation Transfer Function
Complex Aperture OTF:
0 20 40 60 80 100 1200.0
0.2
0.4
0.6
0.8
1.0
MTF
Optics
OQF
Det
Jit
Atm
Smear
Diff
CTE
TOTAL
Freq (mm-1)Data courtesy of Boucher
Note: normalized autocorrelationof the scaled pupil function
R I T Rochester Institute of Technology
Sparse Aperture AutocorrelationSparse Aperture Autocorrelation
• Unaberrated System Pupil:
• Optical Transfer Function (OTF):
[ ] [ ] [ ]∑∑==
−−=−−∗=N
iii
N
iii yyxxsyyxxyxsyxp
11,,,],[ δ
[ ] [ ] [ ]ηλξληλξληξ 22*
22 , ,, zzpzzp −−−−∝ 9H
[ ] [ ] [ ]( ) [ ]∑∑= = =
=−+−+∗−−−−∝N
j
N
k zyzxkjkj yyyxxxyxsyxs
1 1
*
22,,,,ηλξλδηξ 9 H
Pupil
x
y
x1-x1
-d = -(x12+y1
2)1/2
y1
-y1
cross-correlationpairss[x,y]
p[x,y]OTF
ξ
η
y1
-y1
x1-x1-2x1 2x1
d+y1
-d
d
-d-y1
H[ξ,η]
R I T Rochester Institute of Technology
DiffractionDiffraction--Limited MTF ComparisonLimited MTF Comparison
FilledFill Factor: 1.000
CassegrainFill Factor: 0.942
Golay-6Fill Factor: 0.170
AnnularFill Factor: 0.190
Tri-armFill Factor: 0.170
R I T Rochester Institute of Technology
Polychromatic OTF IssuesPolychromatic OTF Issues
– Traditional approaches resample high-resolution panchromatic imagery
– Spectrally weighted polychromatic MTF computed to apply to gray-world scene
– This technique fails to accurately account for scene spectral content and MTF spectral artifacts
– Direct spectral integration of radiance frequency spectrum and system OTF desirable
[ ]( )
[ ] [ ] ( ) ( ) λλληλτληξληξπ
ηξ dLhcf
FTAS
GGS filln
opt0
FTsource,2intdet
ADC
elecconv ,,,,OTF#4
2,out
freq ∫∞
=
[ ]( )
[ ][ ] [ ] ( ) ( ) ( ) λλληλτληξηξπ
ηξ dLFF
hcf
FTASGG
S filln
opt0
sourcepolygrayobj,
grayobj,2
intdet
ADC
elecconvoutfreq ,MTF
0,0,
#42
, ∫∞
=
[ ][ ] ( ) ( ) ( )
( ) ( ) ( )∫
∫=
max
min
max
min
optsource
optsource
poly
,,MTF
,MTF λ
λ
λ
λ
λλληλτλ
λλληλτλληξ
ηξdL
dL
R I T Rochester Institute of Technology
Spectral MTF VariationSpectral MTF VariationComparison of Filled versus Sparse Aperture MTFs
Sparse Aperture MTF:Aberrated 0.24-waves rms
Filled Aperture MTF:Unaberrated
• Oscillatory behavior of sparse aperture MTF versus wavelength creates concern over spectral artifact image quality issues
R I T Rochester Institute of Technology
Simulation OverviewSimulation Overview
Top-level simulation methodology:– Develop geometric pupil function
– Add phase via Zernike polynomials to incorporate aberrations
– Evaluate aberrated system OTF
– Perform Fourier Optical prediction of noise-free degraded image
– Add uncorrelated noise
– Restore imagery via conventional Wiener-Helstrom filter
Original Object
Wiener Filter Restoration
Sparse Aperture PredictionAberrated Sparse Aperture
R I T Rochester Institute of Technology
MTF EvaluationMTF Evaluation(Effects of Aberrations)(Effects of Aberrations)
0.07-wave rms MTFSparse Aperture w/Piston Error
0.10-wave rms MTF 0.20-wave rms MTF
Diffraction-Limited MTF
0.14-wave rms MTF
R I T Rochester Institute of Technology
Modulation Transfer FunctionModulation Transfer Function
Comparison of Filled versus Sparse Aperture MTFs
Aberrated Sparse Aperture MTF: 0.20-wave rms
R I T Rochester Institute of Technology
Point Spread FunctionPoint Spread Function
Comparison of Filled versus Sparse Aperture PSFs
Sparse Aperture PSF:Aberrated 0.20-waves rms
Filled Aperture PSF:Unaberrated
Sparse Aperture PSF:Unaberrated
R I T Rochester Institute of Technology
Quality ImpactsQuality Impacts
Image Quality Comparisonof Filled versus Sparse Aperture
Sparse Aperture:Aberrated 0.20-waves rms
Filled Aperture:Unaberrated
Sparse Aperture:Unaberrated
R I T Rochester Institute of Technology
Wiener Filter ExampleWiener Filter Example
WFE: 0.20-waves rms
Wiener Filtered Image
SNR: 18.5 db
Original Object Noisy Degraded Image
Aperture Comparison Wiener Filtered ImageNoisy Degraded Image
Sparse Aperture Sparse Aperture
Filled Aperture Filled Aperture
Fill Factor: 0.173
SNR: 6.0 db
R I T Rochester Institute of Technology
Image Restoration Example (1)Image Restoration Example (1)Wavefront Error: 0.10-waves rms PTT
Restored RGB Image
SNR: 25.3 db
Sparse Aperturew/Piston Tip/Tilt Errors
Original Object Predicted RGB Image
Aperture Phase Profile Aberrated MTF
Fill Factor: 0.173
R I T Rochester Institute of Technology
Image Restoration Example (2)Image Restoration Example (2)Wavefront Error: 0.20-waves rms PTT
Restored RGB Image
SNR: 25.3 db
Sparse Aperturew/Piston Tip/Tilt Errors
Original Object Predicted RGB Image
Aperture Phase Profile Aberrated MTF
Fill Factor: 0.173
R I T Rochester Institute of Technology
Image Restoration Example (3)Image Restoration Example (3)Wavefront Error: 0.25-waves rms PTT
Restored RGB Image
SNR: 25.3 db
Sparse Aperturew/Piston Tip/Tilt Errors
Original Object Predicted RGB Image
Aperture Phase Profile Aberrated MTF
Fill Factor: 0.173
R I T Rochester Institute of Technology
Closer Examination of RestorationCloser Examination of Restoration
Original Object
R I T Rochester Institute of Technology
Image Restoration Example (1)Image Restoration Example (1)Wavefront Error: 0.10-waves rms PTT SNR: 25.3 dbFill Factor: 0.173
Note correlated noise amplification & presence of color artifacts due to Wiener
filter frequency boost mismatch
Precise knowledge of phase errors assumed
Individual spectral bands restored via central wavelength OTF to simulate physics associated with wideband collection
R I T Rochester Institute of Technology
Image Restoration Example (2)Image Restoration Example (2)Wavefront Error: 0.20-waves rms PTT SNR: 25.3 dbFill Factor: 0.173
Note correlated noise amplification & presence of color artifacts due to Wiener
filter frequency boost mismatch
Precise knowledge of phase errors assumed
Individual spectral bands restored via central wavelength OTF to simulate physics associated with wideband collection
R I T Rochester Institute of Technology
Image Restoration Example (3)Image Restoration Example (3)Wavefront Error: 0.25-waves rms PTT SNR: 25.3 dbFill Factor: 0.173
Note correlated noise amplification & presence of color artifacts due to Wiener
filter frequency boost mismatch
Precise knowledge of phase errors assumed
Individual spectral bands restored via central wavelength OTF to simulate physics associated with wideband collection
R I T Rochester Institute of Technology
Image Restoration Example (4)Image Restoration Example (4)Wavefront Error: 0.30-waves rms defocus
Restored RGB Image
SNR: 16.5 db
Filled Aperturew/Focus Errors
Original Object Predicted RGB Image
Aperture Phase Profile Aberrated MTF
Fill Factor: 1.000
R I T Rochester Institute of Technology
Criticality of Phase KnowledgeCriticality of Phase KnowledgeWavefront Error: 0.25-wave rms PTT
Restored RGB Image(RGB OTF - perfect phase knowledge)
Restored RGB Image(RGB OTF - no phase knowledge)
Restored RGB Image(central OTF - perfect phase knowledge)
SNR: 25.3 db
Sparse Aperturew/Piston Tip/Tilt Errors
Original Object Predicted Noisy ImageAperture Phase Profile
Fill Factor: 0.173
R I T Rochester Institute of Technology
SummarySummary
• Strawman spectra-radiometric sparse aperture model demonstrated
• Includes effects of spectrally-dependent OTF, phase errors, noise, spectral radiance
• Moderately aberrated, low-fill factor sparse aperture systems appear to demonstrate spectral issues