SlideShare a Scribd company logo
1 of 50
Download to read offline
Compressive Spectral Image
Sensing and Optimization
Gonzalo R. Arce
Charles Black Evans Professor
University of Delaware
Newark, Delaware, USA 19716
Distinguished Lecturer Series
Aristotle University of Thessaloniki
March 14, 2014
Optical imaging and spectroscopy discovers the characteristics of scenes and
materials by capturing EM radiation in the 0.01 to 10000 nm spectrum window.
Sensitive not only to spatial and spectral information of a scene, but also to
polarization, tomographic, angular, and even chemical composition.
Multidimensional imaging provides dimension preserving mappings
y = Hf,
where H is the “forward model“, characterizing the focal plane data.
Optical coding shapes H, under some criteria, to match the computational tools
of inverse problems to significantly improve the overall imaging performance.
MURA Hadamard SPECT
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 2 / 50
THE SPECTRAL IMAGING PROBLEM?
Push broom spectral imaging: Traditional approach, expensive, low
sensing speed, senses N × N × L voxels
Optical Filters; Again senses N × N × L voxels; limited by number of
colors
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 3 / 50
Compressive Spectral Imaging (CASSI), New revolutionary method, CS
makes a significant difference, senses only N2
N × N × L
Coded apertures are
the only variable
element.
Coded Apertures are
the key elements in
CASSI.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 4 / 50
WHY IS THIS IMPORTANT?
Remote sensing and surveillance
Visible, NIR, SWIR
Devices are challenging in NIR and SWIR: cost, size,
resolution, cooling
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 5 / 50
WHY IS THIS IMPORTANT?
Medical Imaging: Vascular tissue imaging, angiography,
contrast agent
paint restoration
Compressive Spectral Imaging
Reduce sensing complexity
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 6 / 50
Introduction
Compressive sensing introduced by Donoho†, Candès‡, Tao,
Romberg...
Measurements are given by y = Φx
y
xΦ
M x 1
Measurements
M x N
Sampling Operator
N x 1
Sparse Signal
A sparse solution x is recovered from y by solving the inverse
problem
ˆx = min
x
x 1 s.t. y = Φx.
†
Donoho. IEEE Trans. on Information Theory. December 2006.
‡
Candès, Romberg and Tao. IEEE Trans. on Information Theory. April 2006.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 7 / 50
Introduction
Measurements are given by y = Φx
y ΨΦ α
x
A sparse solution α is recovered from y by solving the inverse
problem
ˆα = min
α
α 1 s.t. y = ΦΨα.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 8 / 50
Introduction
Datacube
f = Ψθ
Compressive Measurements
g = HΨθ + w
Underdetermined system of equations
ˆf = Ψ{min
θ
g − HΨθ 2 + τ θ 1}
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 9 / 50
Coded-Aperture Spectral Imaging (CASSI)
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 10 / 50
Coded-Aperture Spectral Imaging (CASSI)
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 11 / 50
Undetermined system of equations: N × M × L Unknowns and
N(M + L − 1) Equations.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 12 / 50
Computational Model
A single shot compressive measurement across the FPA:
Gnm =
L−1
i=0
Fi(n+m)mTi(n+m) + win
F is the N × M × L datacube
T is the binary code aperture
w is the sensing noise
In vector form, the FPA measurement can be written as
g = Hf + w
H accounts for the aperture code and the dispersive element.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 13 / 50
CASSI Multishot Matrix Model





g0
g1
...
gk−1





=





H0
H1
...
Hk−1





f, (1)
g = Hf, (2)
where H ∈ {0, 1}N(M+L−1)K×NML
.
Multi-shot coding done by using multiple coded apertures or a
Digital-Micromirror-Device (DMD)
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 14 / 50
Matrix CASSI representation g = Hf
Data cube:
N × N × L
Spectral bands: L
Spatial resolution:
N × N
Sensor size
N × (N + L − 1)
V=N(N+L-1)
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 15 / 50
Coded Aperture Optimization: Sensing and
Reconstruction
Restricted Isometry Property in CASSI
Given
g = HΨ
A
θ with |θθθ| = S
The RIP of the CASSI matrix A is defined as the smallest constant δs
such that
(1 − δs) ||θθθ||2
2 ≤ ||Aθθθ||2
2 ≤ (1 + δs) ||θθθ||2
2, (3)
where
δs = max
T⊂[n],|T|≤S
λmax A|T||T| − I (4)
A|T||T| = AT
|T|A|T|, A|T| is a m × |T| matrix whose columns are equal to
|T| columns of the CASSI matrix A, and λmax (.) denotes the largest
eigenvalue.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 16 / 50
Let the entries of ΨΨΨ be Ψj,k and let the columns of ΨΨΨ be [ψψψ0, . . . ,ψψψn−1].
The entries of A|T| can be written as
A|T| jk
= (HψψψΩk
)j = hT
j ψψψΩk
=
L−1
r=0
ti
j−rN
Ψj+r(N ),Ωk
for j = 0, . . . , m − 1, k = 0, . . . , |T| − 1, where i = j/V , N = N2 − N,
and Ωk ∈ {0, . . . , n − 1}. The entries of A|T||T| can be expressed as
A|T||T| jk
=
K−1
i=0
V−1
=0
L−1
r=0
L−1
u=0
ti
−rN
ti
−uN
Ψ +rN ,Ωj
Ψ +uN ,Ωk
(5)
for j, k = 0, . . . , |T| − 1.
Note that the coded aperture products ti
−rN
ti
−uN
determine the
eigenvalues of A|T||T|, and consequently they determine the constant
δs in the RIP.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 17 / 50
Boolean Coded Apertures
An optimal ensemble of four 64 × 64 boolean coded apertures.
Each spatial
coordinate in the
ensemble contains
only one 1-value
entry.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 18 / 50
Bernoulli Coded Apertures
An ensemble of four 64 × 64 Bernoulli coded apertures.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 19 / 50
Performance of coded apertures
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 20 / 50
Original datacube Boolean (40.4dB) Unsigned grayscale (31.2dB)
Binary (27.7dB) Hadamard (27.7dB) Signed grayscale (22.7dB)
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 21 / 50
Original datacube Boolean Unsigned grayscale
Binary Hadamard Signed grayscale
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 22 / 50
Coded Aperture Optimization: Spectral Selectivity
UAV sensor requirements depend on flight duration, range,
altitude, etc.
Need: Hyperspectral imaging that dynamically adapts to optimal
spectral bands.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 23 / 50
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 24 / 50
((a)) Original ((b)) 12 Random Codes ((c)) 9 Optimal Codes
The resulting spectral data cubes are shown as they would be viewed by a Stingray
F-033C CCD Color Camera.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 25 / 50
((a)) Random Code ((b)) Original ((c)) Optimal Code
((d)) Random Codes ((e)) Optimized Codes
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 26 / 50
((a)) Random ((b)) Optimized
((c)) ((d))
Differences between the original and the reconstructed 3rd
spectral channel (479nm)
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 27 / 50
((a)) Original ((b)) 12 Random Codes ((c)) 12 Optimized Codes
The resulting spectral data cubes are shown as they would be viewed by a Stingray
F-033C CCD Color Camera.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 28 / 50
Coded Aperture Optimization: Image Classification
Goal: classification of a spectral scene using
Compressive measurements
Optimal code aperture
Sparsity signal model
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 29 / 50
Given the FPA measurements g, every test pixel fi belongs to one of
the P known classes
{H(1)
, H(2)
, ..., H(P)
}
0 10 20 30 40 50 60 70
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Spectrum Band
Glutamine
0 10 20 30 40 50 60 70
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Spectrum Band
Histidine
0 10 20 30 40 50 60 70
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Spectrum Band
Isoleucine
If a pixel fi ∈ H(k), its spectral profile lies in a low-dimensional
subspace spanned by the training samples: {s
(k)
j }j=1,...,Np
fi ≈ [s
(k)
1 , s
(k)
2 , ..., s
(k)
Nk
][α
(k)
1 , α
(k)
2 , ..., α
(k)
Np
]T
= S(k)
α(k)
where, α(k) is a sparse vector.Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 30 / 50
Sparsity Model
Combining all class sub-dictionaries
fi ≈ [S(1)
, ..., S(k)
, ..., S(P)
][α(1)
, ..., α(k)
, ..., α(P)
]T
= Sα
Ideally, if fi ∈ H(k), then
α(j) = 0; ∀j = 1, ..., P; j = k.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 31 / 50
Superposition of 3 shots for the different codes
Optimal codes
5 10 15 20 25 30
5
10
15
20
25
30
0
0.5
1
1.5
2
2.5
3
Hadamard Codes
5 10 15 20 25 30
5
10
15
20
25
30
0
0.5
1
1.5
2
2.5
3
Bernoulli codes
5 10 15 20 25 30
5
10
15
20
25
30
0
0.5
1
1.5
2
2.5
3
Ck = K−1
k=0 (tk
m,n)2 at each (m, n) spatial location.
Optimal Hadamard Bernoulli
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 32 / 50
Image Classification Results
200 spectral bands. 50 shots.
Ground-truth Single spectral band SVM-full datacube (73.5%)
Bernoulli-25%datacube
(64.3%)
Hadamard-25%datacube
(63.4%)
Optimal-25%datacube (73.7%)
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 33 / 50
Image Classification Results
200 spectral bands. 50 shots.
Ground-truth FPA measurement SVM-full datacube (73.5%)
Bernoulli-25%datacube
(64.3%)
Hadamard-25%datacube
(63.4%)
Optimal-25%datacube (73.7%)
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 34 / 50
Colored Coded Aperture Spectral Imaging
Patterned coating combines micro-lithography with optical coating
technology.
Precision patterned coating
and patterns
Sub-pixel alignment accuracy
Ultraviolet, visible, NIR, SWIR
Multi-filter arrays on monolithic
substrates
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 35 / 50
NEW FAMILY OF CODED APERTURES
Boolean Spectrally Selective
Super-resolution Colored
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 36 / 50
Colored coded aperture model
Colored coded aperture is a color filter array
Each entry is a wavelength selective color filter
3D Mask model has the same dimensions than the objective discrete data cube
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 37 / 50
Linear dispersion and focal plane array integration
Linear shifting operation
Focal plane array (FPA) projections
The number of pixels of the FPA
detector is N(N + L − 1) N2
L
(size of the spectral data cube)
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 38 / 50
Random Boolean Code
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 39 / 50
4 Colors Random Code
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 40 / 50
Restricted Isometry Property of Colored CASSI
A = HΨΨΨ, fff = ΨΨΨθθθ, ΨΨΨ = W ⊗ ΨΨΨ2D
Definition
(1 − δs) ||θθθ||2
2 ≤ ||Aθθθ||2
2 ≤ (1 + δs) ||θθθ||2
2,
δs = max
T ⊂[N2L],|T |≤S
||AT
|T |A|T | − I||2
2,
A|T |, |T | columns of A indexed by the set T
δs = max
T ⊂[N2L],|T |≤S
λmax A|T ||T | − I
A|T | ir
= hiψψψΩr
=
L−1
k=0
t i
k mi −kN
Ψmi +k(N ),Ωr
(hi )j =



t i
kj i− i V−kj N
, if i − i V = j − kj N
0, otherwise,
A = HΨΨΨ2D
A = H W ⊗ ΨΨΨ2D
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 41 / 50
A|T ||T | = AT
|T |A|T |
A|T ||T | r,u
=
K−1
=0
V−1
i=0
L−1
k1=0
L−1
k2=0
tk1
i−k1N
tk2
i−k2N
Ψi+k1N ,ΩrΨi+k2N ,Ωu
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 42 / 50
Results: LH-Colored Coded Aperture
A geometric interpretation of the colored coded apertures for LH-Colored filters (3
shots).
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 43 / 50
Code Design Results B-Colored Coded Apertures (Band Pass Filters)
Geometric interpretation of colored coded apertures for B-filters (3 shots)
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 44 / 50
461nm 470nm 479nm 488nm
497nm 506nm 515nm 524nm
533nm 542nm 551nm 560nm
569nm 578nm 587nm 596nm
16 channels
461-596nm
256 × 256
pixels
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 45 / 50
Reconstruction From B-Colored Coded Apertures
2 4 6 8 10 12 14
30
40
50
60
70
80
90
Number of Shots
PSNR Block Unblock
Random B-colored
B-colored
Mean PSNR of the reconstructed data cubes.
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 46 / 50
Reconstructions from Real Measurements (RGB)
K = 1 K = 1
Sh=1
(i) 1 snapshot: Photomask, Coloring, Optimal
K = 4 K = 4
Sh=4
(j) 4 snapshots: Photoask, Coloring, Optimal
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 47 / 50
Reconstruction Results (Photomask vs Colored
Coded Apertures)
454 nm K = 1 468 nm K = 1 485 nm K = 1 506 nm K = 1 529 nm K = 1 557 nm K = 1 594 nm K = 1 639 nm K = 1
454 nm K = 1 468 nm K = 1 485 nm K = 1 506 nm K = 1 529 nm K = 1 557 nm K = 1 594 nm K = 1 639 nm K = 1
(a) One snapshot reconstruction. (First row) Photomask, (Second row) Coloring
454 nm K = 4 468 nm K = 4 485 nm K = 4 506 nm K = 4 529 nm K = 4 557 nm K = 4 594 nm K = 4 639 nm K = 4
454 nm K = 4 468 nm K = 4 485 nm K = 4 506 nm K = 4 529 nm K = 4 557 nm K = 4 594 nm K = 4 639 nm K = 4
(b) 4 snapshots reconstructions. (First row) Photomask, (Second row) Coloring
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 48 / 50
Spectral Reconstruction From Colored Coded Apertures
Sh=4
P1
P2
454 468 485 506 529 557 594 639
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Nanometers
Norm.Intensity
Spectral Comparison: 4 Snapshots
Spectrometer
Photomask
Coloring
Optimal
454 468 485 506 529 557 594 639
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Nanometers
Norm.Intensity
Spectral Comparison: 4 Snapshots
Spectrometer
Photomask
Coloring
Optimal
Pixel P1
Pixel P2
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 49 / 50
Conclusions
Optical Coding for Compressive Spectral Imaging
Good codes for reconstruction, classification, unmixing
Colored codes offer multidimensional coding - open problem
Convolution Optical Coding (Projections)
Light field imaging
X-ray tomography
Super-resolution microscopy
Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 50 / 50

More Related Content

What's hot

Image Restitution Using Non-Locally Centralized Sparse Representation Model
Image Restitution Using Non-Locally Centralized Sparse Representation ModelImage Restitution Using Non-Locally Centralized Sparse Representation Model
Image Restitution Using Non-Locally Centralized Sparse Representation ModelIJERA Editor
 
U tokyo 2019
U tokyo 2019U tokyo 2019
U tokyo 2019Jinze Yu
 
Macrocanonical models for texture synthesis
Macrocanonical models for texture synthesisMacrocanonical models for texture synthesis
Macrocanonical models for texture synthesisValentin De Bortoli
 
Elementary Landscape Decomposition of the Test Suite Minimization Problem
Elementary Landscape Decomposition of the Test Suite Minimization ProblemElementary Landscape Decomposition of the Test Suite Minimization Problem
Elementary Landscape Decomposition of the Test Suite Minimization Problemjfrchicanog
 
Thresholding eqns for wavelet
Thresholding eqns for waveletThresholding eqns for wavelet
Thresholding eqns for waveletajayhakkumar
 
A Compressed Sensing Approach to Image Reconstruction
A Compressed Sensing Approach to Image ReconstructionA Compressed Sensing Approach to Image Reconstruction
A Compressed Sensing Approach to Image Reconstructionijsrd.com
 
iDiff 2008 conference #04 IP-Racine FSSG
iDiff 2008 conference #04 IP-Racine   FSSGiDiff 2008 conference #04 IP-Racine   FSSG
iDiff 2008 conference #04 IP-Racine FSSGBenoit Michel
 
Quantitative Propagation of Chaos for SGD in Wide Neural Networks
Quantitative Propagation of Chaos for SGD in Wide Neural NetworksQuantitative Propagation of Chaos for SGD in Wide Neural Networks
Quantitative Propagation of Chaos for SGD in Wide Neural NetworksValentin De Bortoli
 
Kernel-Based_Retrieval_of_Atmospheric_Profiles_from_IASI_Data.pdf
Kernel-Based_Retrieval_of_Atmospheric_Profiles_from_IASI_Data.pdfKernel-Based_Retrieval_of_Atmospheric_Profiles_from_IASI_Data.pdf
Kernel-Based_Retrieval_of_Atmospheric_Profiles_from_IASI_Data.pdfgrssieee
 
Good denoising using wavelets
Good denoising using waveletsGood denoising using wavelets
Good denoising using waveletsbeenamohan
 
Mesh Generation and Topological Data Analysis
Mesh Generation and Topological Data AnalysisMesh Generation and Topological Data Analysis
Mesh Generation and Topological Data AnalysisDon Sheehy
 
Real Time Localization Using Receiver Signal Strength Indicator
Real Time Localization Using Receiver Signal Strength IndicatorReal Time Localization Using Receiver Signal Strength Indicator
Real Time Localization Using Receiver Signal Strength IndicatorRana Basheer
 
05_글로벌일루미네이션
05_글로벌일루미네이션05_글로벌일루미네이션
05_글로벌일루미네이션noerror
 
Image Denoising Using Wavelet
Image Denoising Using WaveletImage Denoising Using Wavelet
Image Denoising Using WaveletAsim Qureshi
 
International Journal of Engineering Research and Development (IJERD)
 International Journal of Engineering Research and Development (IJERD) International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)IJERD Editor
 

What's hot (20)

Nonlinear observer design
Nonlinear observer designNonlinear observer design
Nonlinear observer design
 
Image Restitution Using Non-Locally Centralized Sparse Representation Model
Image Restitution Using Non-Locally Centralized Sparse Representation ModelImage Restitution Using Non-Locally Centralized Sparse Representation Model
Image Restitution Using Non-Locally Centralized Sparse Representation Model
 
U tokyo 2019
U tokyo 2019U tokyo 2019
U tokyo 2019
 
190 195
190 195190 195
190 195
 
Macrocanonical models for texture synthesis
Macrocanonical models for texture synthesisMacrocanonical models for texture synthesis
Macrocanonical models for texture synthesis
 
Ax26326329
Ax26326329Ax26326329
Ax26326329
 
Gtti 10032021
Gtti 10032021Gtti 10032021
Gtti 10032021
 
Elementary Landscape Decomposition of the Test Suite Minimization Problem
Elementary Landscape Decomposition of the Test Suite Minimization ProblemElementary Landscape Decomposition of the Test Suite Minimization Problem
Elementary Landscape Decomposition of the Test Suite Minimization Problem
 
Thresholding eqns for wavelet
Thresholding eqns for waveletThresholding eqns for wavelet
Thresholding eqns for wavelet
 
A Compressed Sensing Approach to Image Reconstruction
A Compressed Sensing Approach to Image ReconstructionA Compressed Sensing Approach to Image Reconstruction
A Compressed Sensing Approach to Image Reconstruction
 
iDiff 2008 conference #04 IP-Racine FSSG
iDiff 2008 conference #04 IP-Racine   FSSGiDiff 2008 conference #04 IP-Racine   FSSG
iDiff 2008 conference #04 IP-Racine FSSG
 
Quantitative Propagation of Chaos for SGD in Wide Neural Networks
Quantitative Propagation of Chaos for SGD in Wide Neural NetworksQuantitative Propagation of Chaos for SGD in Wide Neural Networks
Quantitative Propagation of Chaos for SGD in Wide Neural Networks
 
Kernel-Based_Retrieval_of_Atmospheric_Profiles_from_IASI_Data.pdf
Kernel-Based_Retrieval_of_Atmospheric_Profiles_from_IASI_Data.pdfKernel-Based_Retrieval_of_Atmospheric_Profiles_from_IASI_Data.pdf
Kernel-Based_Retrieval_of_Atmospheric_Profiles_from_IASI_Data.pdf
 
Good denoising using wavelets
Good denoising using waveletsGood denoising using wavelets
Good denoising using wavelets
 
Mesh Generation and Topological Data Analysis
Mesh Generation and Topological Data AnalysisMesh Generation and Topological Data Analysis
Mesh Generation and Topological Data Analysis
 
Hr3114661470
Hr3114661470Hr3114661470
Hr3114661470
 
Real Time Localization Using Receiver Signal Strength Indicator
Real Time Localization Using Receiver Signal Strength IndicatorReal Time Localization Using Receiver Signal Strength Indicator
Real Time Localization Using Receiver Signal Strength Indicator
 
05_글로벌일루미네이션
05_글로벌일루미네이션05_글로벌일루미네이션
05_글로벌일루미네이션
 
Image Denoising Using Wavelet
Image Denoising Using WaveletImage Denoising Using Wavelet
Image Denoising Using Wavelet
 
International Journal of Engineering Research and Development (IJERD)
 International Journal of Engineering Research and Development (IJERD) International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)
 

Viewers also liked

Viewers also liked (20)

Influence Propagation in Large Graphs - Theorems and Algorithms
Influence Propagation in Large Graphs - Theorems and AlgorithmsInfluence Propagation in Large Graphs - Theorems and Algorithms
Influence Propagation in Large Graphs - Theorems and Algorithms
 
Sparse and Low Rank Representations in Music Signal Analysis
 Sparse and Low Rank Representations in Music Signal  Analysis Sparse and Low Rank Representations in Music Signal  Analysis
Sparse and Low Rank Representations in Music Signal Analysis
 
Tribute to Nicolas Galatsanos
Tribute to Nicolas GalatsanosTribute to Nicolas Galatsanos
Tribute to Nicolas Galatsanos
 
Web Usage Miningand Using Ontology for Capturing Web Usage Semantic
Web Usage Miningand Using Ontology for Capturing Web Usage SemanticWeb Usage Miningand Using Ontology for Capturing Web Usage Semantic
Web Usage Miningand Using Ontology for Capturing Web Usage Semantic
 
Data Quality: Not Your Typical Database Problem
Data Quality: Not Your Typical Database ProblemData Quality: Not Your Typical Database Problem
Data Quality: Not Your Typical Database Problem
 
Co-evolution, Games, and Social Behaviors
Co-evolution, Games, and Social BehaviorsCo-evolution, Games, and Social Behaviors
Co-evolution, Games, and Social Behaviors
 
Sparsity Control for Robustness and Social Data Analysis
Sparsity Control for Robustness and Social Data AnalysisSparsity Control for Robustness and Social Data Analysis
Sparsity Control for Robustness and Social Data Analysis
 
From Programs to Systems – Building a Smarter World
From Programs to Systems – Building a Smarter WorldFrom Programs to Systems – Building a Smarter World
From Programs to Systems – Building a Smarter World
 
A Classification Framework For Component Models
 A Classification Framework For Component Models A Classification Framework For Component Models
A Classification Framework For Component Models
 
Opening Second Greek Signal Processing Jam
Opening Second Greek Signal Processing JamOpening Second Greek Signal Processing Jam
Opening Second Greek Signal Processing Jam
 
The Tower of Knowledge A Generic System Architecture
The Tower of Knowledge A Generic System ArchitectureThe Tower of Knowledge A Generic System Architecture
The Tower of Knowledge A Generic System Architecture
 
Nonlinear Communications: Achievable Rates, Estimation, and Decoding
Nonlinear Communications: Achievable Rates, Estimation, and DecodingNonlinear Communications: Achievable Rates, Estimation, and Decoding
Nonlinear Communications: Achievable Rates, Estimation, and Decoding
 
State Space Exploration for NASA’s Safety Critical Systems
State Space Exploration for NASA’s Safety Critical SystemsState Space Exploration for NASA’s Safety Critical Systems
State Space Exploration for NASA’s Safety Critical Systems
 
Semantic 3DTV Content Analysis and Description
Semantic 3DTV Content Analysis and DescriptionSemantic 3DTV Content Analysis and Description
Semantic 3DTV Content Analysis and Description
 
Jamming in Wireless Sensor Networks
Jamming in Wireless Sensor NetworksJamming in Wireless Sensor Networks
Jamming in Wireless Sensor Networks
 
Mixture Models for Image Analysis
Mixture Models for Image AnalysisMixture Models for Image Analysis
Mixture Models for Image Analysis
 
Sparse and Redundant Representations: Theory and Applications
Sparse and Redundant Representations: Theory and ApplicationsSparse and Redundant Representations: Theory and Applications
Sparse and Redundant Representations: Theory and Applications
 
Networked 3-D Virtual Collaboration in Science and Education: Towards 'Web 3....
Networked 3-D Virtual Collaboration in Science and Education: Towards 'Web 3....Networked 3-D Virtual Collaboration in Science and Education: Towards 'Web 3....
Networked 3-D Virtual Collaboration in Science and Education: Towards 'Web 3....
 
Machine Learning Tools and Particle Swarm Optimization for Content-Based Sear...
Machine Learning Tools and Particle Swarm Optimization for Content-Based Sear...Machine Learning Tools and Particle Swarm Optimization for Content-Based Sear...
Machine Learning Tools and Particle Swarm Optimization for Content-Based Sear...
 
Artificial Intelligence and Human Thinking
Artificial Intelligence and Human ThinkingArtificial Intelligence and Human Thinking
Artificial Intelligence and Human Thinking
 

Similar to Compressive Spectral Image Sensing, Processing, and Optimization

ALEXANDER FRACTIONAL INTEGRAL FILTERING OF WAVELET COEFFICIENTS FOR IMAGE DEN...
ALEXANDER FRACTIONAL INTEGRAL FILTERING OF WAVELET COEFFICIENTS FOR IMAGE DEN...ALEXANDER FRACTIONAL INTEGRAL FILTERING OF WAVELET COEFFICIENTS FOR IMAGE DEN...
ALEXANDER FRACTIONAL INTEGRAL FILTERING OF WAVELET COEFFICIENTS FOR IMAGE DEN...sipij
 
Slides: On the Chi Square and Higher-Order Chi Distances for Approximating f-...
Slides: On the Chi Square and Higher-Order Chi Distances for Approximating f-...Slides: On the Chi Square and Higher-Order Chi Distances for Approximating f-...
Slides: On the Chi Square and Higher-Order Chi Distances for Approximating f-...Frank Nielsen
 
Image formation
Image formationImage formation
Image formationpotaters
 
ABC with Wasserstein distances
ABC with Wasserstein distancesABC with Wasserstein distances
ABC with Wasserstein distancesChristian Robert
 
A MODIFIED DIRECTIONAL WEIGHTED CASCADED-MASK MEDIAN FILTER FOR REMOVAL OF RA...
A MODIFIED DIRECTIONAL WEIGHTED CASCADED-MASK MEDIAN FILTER FOR REMOVAL OF RA...A MODIFIED DIRECTIONAL WEIGHTED CASCADED-MASK MEDIAN FILTER FOR REMOVAL OF RA...
A MODIFIED DIRECTIONAL WEIGHTED CASCADED-MASK MEDIAN FILTER FOR REMOVAL OF RA...cscpconf
 
On the smallest enclosing information disk
 On the smallest enclosing information disk On the smallest enclosing information disk
On the smallest enclosing information diskFrank Nielsen
 
Slides: A glance at information-geometric signal processing
Slides: A glance at information-geometric signal processingSlides: A glance at information-geometric signal processing
Slides: A glance at information-geometric signal processingFrank Nielsen
 
ABC based on Wasserstein distances
ABC based on Wasserstein distancesABC based on Wasserstein distances
ABC based on Wasserstein distancesChristian Robert
 
Patch Matching with Polynomial Exponential Families and Projective Divergences
Patch Matching with Polynomial Exponential Families and Projective DivergencesPatch Matching with Polynomial Exponential Families and Projective Divergences
Patch Matching with Polynomial Exponential Families and Projective DivergencesFrank Nielsen
 
Computational Information Geometry on Matrix Manifolds (ICTP 2013)
Computational Information Geometry on Matrix Manifolds (ICTP 2013)Computational Information Geometry on Matrix Manifolds (ICTP 2013)
Computational Information Geometry on Matrix Manifolds (ICTP 2013)Frank Nielsen
 
LightFields.jl: Fast 3D image reconstruction for VR applications - Hector And...
LightFields.jl: Fast 3D image reconstruction for VR applications - Hector And...LightFields.jl: Fast 3D image reconstruction for VR applications - Hector And...
LightFields.jl: Fast 3D image reconstruction for VR applications - Hector And...PyData
 
A New Enhanced Method of Non Parametric power spectrum Estimation.
A New Enhanced Method of Non Parametric power spectrum Estimation.A New Enhanced Method of Non Parametric power spectrum Estimation.
A New Enhanced Method of Non Parametric power spectrum Estimation.CSCJournals
 
Detection of unknown signal
Detection of unknown signalDetection of unknown signal
Detection of unknown signalsumitf1
 
Color Img at Prisma Network meeting 2009
Color Img at Prisma Network meeting 2009Color Img at Prisma Network meeting 2009
Color Img at Prisma Network meeting 2009Juan Luis Nieves
 
Research Inventy : International Journal of Engineering and Science is publis...
Research Inventy : International Journal of Engineering and Science is publis...Research Inventy : International Journal of Engineering and Science is publis...
Research Inventy : International Journal of Engineering and Science is publis...researchinventy
 
Blind Beam-Hardening Correction from Poisson Measurements
Blind Beam-Hardening Correction from Poisson MeasurementsBlind Beam-Hardening Correction from Poisson Measurements
Blind Beam-Hardening Correction from Poisson MeasurementsAleksandar Dogandžić
 

Similar to Compressive Spectral Image Sensing, Processing, and Optimization (20)

CS-ChapterI.pdf
CS-ChapterI.pdfCS-ChapterI.pdf
CS-ChapterI.pdf
 
ALEXANDER FRACTIONAL INTEGRAL FILTERING OF WAVELET COEFFICIENTS FOR IMAGE DEN...
ALEXANDER FRACTIONAL INTEGRAL FILTERING OF WAVELET COEFFICIENTS FOR IMAGE DEN...ALEXANDER FRACTIONAL INTEGRAL FILTERING OF WAVELET COEFFICIENTS FOR IMAGE DEN...
ALEXANDER FRACTIONAL INTEGRAL FILTERING OF WAVELET COEFFICIENTS FOR IMAGE DEN...
 
Slides: On the Chi Square and Higher-Order Chi Distances for Approximating f-...
Slides: On the Chi Square and Higher-Order Chi Distances for Approximating f-...Slides: On the Chi Square and Higher-Order Chi Distances for Approximating f-...
Slides: On the Chi Square and Higher-Order Chi Distances for Approximating f-...
 
Image formation
Image formationImage formation
Image formation
 
ABC with Wasserstein distances
ABC with Wasserstein distancesABC with Wasserstein distances
ABC with Wasserstein distances
 
A MODIFIED DIRECTIONAL WEIGHTED CASCADED-MASK MEDIAN FILTER FOR REMOVAL OF RA...
A MODIFIED DIRECTIONAL WEIGHTED CASCADED-MASK MEDIAN FILTER FOR REMOVAL OF RA...A MODIFIED DIRECTIONAL WEIGHTED CASCADED-MASK MEDIAN FILTER FOR REMOVAL OF RA...
A MODIFIED DIRECTIONAL WEIGHTED CASCADED-MASK MEDIAN FILTER FOR REMOVAL OF RA...
 
On the smallest enclosing information disk
 On the smallest enclosing information disk On the smallest enclosing information disk
On the smallest enclosing information disk
 
Slides: A glance at information-geometric signal processing
Slides: A glance at information-geometric signal processingSlides: A glance at information-geometric signal processing
Slides: A glance at information-geometric signal processing
 
ABC based on Wasserstein distances
ABC based on Wasserstein distancesABC based on Wasserstein distances
ABC based on Wasserstein distances
 
The LabPQR Color Space
The LabPQR Color SpaceThe LabPQR Color Space
The LabPQR Color Space
 
Patch Matching with Polynomial Exponential Families and Projective Divergences
Patch Matching with Polynomial Exponential Families and Projective DivergencesPatch Matching with Polynomial Exponential Families and Projective Divergences
Patch Matching with Polynomial Exponential Families and Projective Divergences
 
Computational Information Geometry on Matrix Manifolds (ICTP 2013)
Computational Information Geometry on Matrix Manifolds (ICTP 2013)Computational Information Geometry on Matrix Manifolds (ICTP 2013)
Computational Information Geometry on Matrix Manifolds (ICTP 2013)
 
Deblurring in ct
Deblurring in ctDeblurring in ct
Deblurring in ct
 
LightFields.jl: Fast 3D image reconstruction for VR applications - Hector And...
LightFields.jl: Fast 3D image reconstruction for VR applications - Hector And...LightFields.jl: Fast 3D image reconstruction for VR applications - Hector And...
LightFields.jl: Fast 3D image reconstruction for VR applications - Hector And...
 
A New Enhanced Method of Non Parametric power spectrum Estimation.
A New Enhanced Method of Non Parametric power spectrum Estimation.A New Enhanced Method of Non Parametric power spectrum Estimation.
A New Enhanced Method of Non Parametric power spectrum Estimation.
 
Lecture 3 sapienza 2017
Lecture 3 sapienza 2017Lecture 3 sapienza 2017
Lecture 3 sapienza 2017
 
Detection of unknown signal
Detection of unknown signalDetection of unknown signal
Detection of unknown signal
 
Color Img at Prisma Network meeting 2009
Color Img at Prisma Network meeting 2009Color Img at Prisma Network meeting 2009
Color Img at Prisma Network meeting 2009
 
Research Inventy : International Journal of Engineering and Science is publis...
Research Inventy : International Journal of Engineering and Science is publis...Research Inventy : International Journal of Engineering and Science is publis...
Research Inventy : International Journal of Engineering and Science is publis...
 
Blind Beam-Hardening Correction from Poisson Measurements
Blind Beam-Hardening Correction from Poisson MeasurementsBlind Beam-Hardening Correction from Poisson Measurements
Blind Beam-Hardening Correction from Poisson Measurements
 

Recently uploaded

Install Stable Diffusion in windows machine
Install Stable Diffusion in windows machineInstall Stable Diffusion in windows machine
Install Stable Diffusion in windows machinePadma Pradeep
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsEnterprise Knowledge
 
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024BookNet Canada
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonetsnaman860154
 
Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsMaria Levchenko
 
Pigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping ElbowsPigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping ElbowsPigging Solutions
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountPuma Security, LLC
 
08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking MenDelhi Call girls
 
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmaticsKotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmaticscarlostorres15106
 
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | DelhiFULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhisoniya singh
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)Gabriella Davis
 
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024BookNet Canada
 
Human Factors of XR: Using Human Factors to Design XR Systems
Human Factors of XR: Using Human Factors to Design XR SystemsHuman Factors of XR: Using Human Factors to Design XR Systems
Human Factors of XR: Using Human Factors to Design XR SystemsMark Billinghurst
 
SIEMENS: RAPUNZEL – A Tale About Knowledge Graph
SIEMENS: RAPUNZEL – A Tale About Knowledge GraphSIEMENS: RAPUNZEL – A Tale About Knowledge Graph
SIEMENS: RAPUNZEL – A Tale About Knowledge GraphNeo4j
 
Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for PartnersEnhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for PartnersThousandEyes
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationMichael W. Hawkins
 
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...Patryk Bandurski
 
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationBeyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationSafe Software
 
Azure Monitor & Application Insight to monitor Infrastructure & Application
Azure Monitor & Application Insight to monitor Infrastructure & ApplicationAzure Monitor & Application Insight to monitor Infrastructure & Application
Azure Monitor & Application Insight to monitor Infrastructure & ApplicationAndikSusilo4
 
Benefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other FrameworksBenefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other FrameworksSoftradix Technologies
 

Recently uploaded (20)

Install Stable Diffusion in windows machine
Install Stable Diffusion in windows machineInstall Stable Diffusion in windows machine
Install Stable Diffusion in windows machine
 
IAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI SolutionsIAC 2024 - IA Fast Track to Search Focused AI Solutions
IAC 2024 - IA Fast Track to Search Focused AI Solutions
 
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
Transcript: #StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonets
 
Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed texts
 
Pigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping ElbowsPigging Solutions Piggable Sweeping Elbows
Pigging Solutions Piggable Sweeping Elbows
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path Mount
 
08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men08448380779 Call Girls In Friends Colony Women Seeking Men
08448380779 Call Girls In Friends Colony Women Seeking Men
 
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmaticsKotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
 
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | DelhiFULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)
 
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
 
Human Factors of XR: Using Human Factors to Design XR Systems
Human Factors of XR: Using Human Factors to Design XR SystemsHuman Factors of XR: Using Human Factors to Design XR Systems
Human Factors of XR: Using Human Factors to Design XR Systems
 
SIEMENS: RAPUNZEL – A Tale About Knowledge Graph
SIEMENS: RAPUNZEL – A Tale About Knowledge GraphSIEMENS: RAPUNZEL – A Tale About Knowledge Graph
SIEMENS: RAPUNZEL – A Tale About Knowledge Graph
 
Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for PartnersEnhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
 
GenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day PresentationGenCyber Cyber Security Day Presentation
GenCyber Cyber Security Day Presentation
 
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
 
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationBeyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
 
Azure Monitor & Application Insight to monitor Infrastructure & Application
Azure Monitor & Application Insight to monitor Infrastructure & ApplicationAzure Monitor & Application Insight to monitor Infrastructure & Application
Azure Monitor & Application Insight to monitor Infrastructure & Application
 
Benefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other FrameworksBenefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other Frameworks
 

Compressive Spectral Image Sensing, Processing, and Optimization

  • 1. Compressive Spectral Image Sensing and Optimization Gonzalo R. Arce Charles Black Evans Professor University of Delaware Newark, Delaware, USA 19716 Distinguished Lecturer Series Aristotle University of Thessaloniki March 14, 2014
  • 2. Optical imaging and spectroscopy discovers the characteristics of scenes and materials by capturing EM radiation in the 0.01 to 10000 nm spectrum window. Sensitive not only to spatial and spectral information of a scene, but also to polarization, tomographic, angular, and even chemical composition. Multidimensional imaging provides dimension preserving mappings y = Hf, where H is the “forward model“, characterizing the focal plane data. Optical coding shapes H, under some criteria, to match the computational tools of inverse problems to significantly improve the overall imaging performance. MURA Hadamard SPECT Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 2 / 50
  • 3. THE SPECTRAL IMAGING PROBLEM? Push broom spectral imaging: Traditional approach, expensive, low sensing speed, senses N × N × L voxels Optical Filters; Again senses N × N × L voxels; limited by number of colors Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 3 / 50
  • 4. Compressive Spectral Imaging (CASSI), New revolutionary method, CS makes a significant difference, senses only N2 N × N × L Coded apertures are the only variable element. Coded Apertures are the key elements in CASSI. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 4 / 50
  • 5. WHY IS THIS IMPORTANT? Remote sensing and surveillance Visible, NIR, SWIR Devices are challenging in NIR and SWIR: cost, size, resolution, cooling Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 5 / 50
  • 6. WHY IS THIS IMPORTANT? Medical Imaging: Vascular tissue imaging, angiography, contrast agent paint restoration Compressive Spectral Imaging Reduce sensing complexity Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 6 / 50
  • 7. Introduction Compressive sensing introduced by Donoho†, Candès‡, Tao, Romberg... Measurements are given by y = Φx y xΦ M x 1 Measurements M x N Sampling Operator N x 1 Sparse Signal A sparse solution x is recovered from y by solving the inverse problem ˆx = min x x 1 s.t. y = Φx. † Donoho. IEEE Trans. on Information Theory. December 2006. ‡ Candès, Romberg and Tao. IEEE Trans. on Information Theory. April 2006. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 7 / 50
  • 8. Introduction Measurements are given by y = Φx y ΨΦ α x A sparse solution α is recovered from y by solving the inverse problem ˆα = min α α 1 s.t. y = ΦΨα. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 8 / 50
  • 9. Introduction Datacube f = Ψθ Compressive Measurements g = HΨθ + w Underdetermined system of equations ˆf = Ψ{min θ g − HΨθ 2 + τ θ 1} Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 9 / 50
  • 10. Coded-Aperture Spectral Imaging (CASSI) Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 10 / 50
  • 11. Coded-Aperture Spectral Imaging (CASSI) Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 11 / 50
  • 12. Undetermined system of equations: N × M × L Unknowns and N(M + L − 1) Equations. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 12 / 50
  • 13. Computational Model A single shot compressive measurement across the FPA: Gnm = L−1 i=0 Fi(n+m)mTi(n+m) + win F is the N × M × L datacube T is the binary code aperture w is the sensing noise In vector form, the FPA measurement can be written as g = Hf + w H accounts for the aperture code and the dispersive element. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 13 / 50
  • 14. CASSI Multishot Matrix Model      g0 g1 ... gk−1      =      H0 H1 ... Hk−1      f, (1) g = Hf, (2) where H ∈ {0, 1}N(M+L−1)K×NML . Multi-shot coding done by using multiple coded apertures or a Digital-Micromirror-Device (DMD) Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 14 / 50
  • 15. Matrix CASSI representation g = Hf Data cube: N × N × L Spectral bands: L Spatial resolution: N × N Sensor size N × (N + L − 1) V=N(N+L-1) Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 15 / 50
  • 16. Coded Aperture Optimization: Sensing and Reconstruction Restricted Isometry Property in CASSI Given g = HΨ A θ with |θθθ| = S The RIP of the CASSI matrix A is defined as the smallest constant δs such that (1 − δs) ||θθθ||2 2 ≤ ||Aθθθ||2 2 ≤ (1 + δs) ||θθθ||2 2, (3) where δs = max T⊂[n],|T|≤S λmax A|T||T| − I (4) A|T||T| = AT |T|A|T|, A|T| is a m × |T| matrix whose columns are equal to |T| columns of the CASSI matrix A, and λmax (.) denotes the largest eigenvalue. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 16 / 50
  • 17. Let the entries of ΨΨΨ be Ψj,k and let the columns of ΨΨΨ be [ψψψ0, . . . ,ψψψn−1]. The entries of A|T| can be written as A|T| jk = (HψψψΩk )j = hT j ψψψΩk = L−1 r=0 ti j−rN Ψj+r(N ),Ωk for j = 0, . . . , m − 1, k = 0, . . . , |T| − 1, where i = j/V , N = N2 − N, and Ωk ∈ {0, . . . , n − 1}. The entries of A|T||T| can be expressed as A|T||T| jk = K−1 i=0 V−1 =0 L−1 r=0 L−1 u=0 ti −rN ti −uN Ψ +rN ,Ωj Ψ +uN ,Ωk (5) for j, k = 0, . . . , |T| − 1. Note that the coded aperture products ti −rN ti −uN determine the eigenvalues of A|T||T|, and consequently they determine the constant δs in the RIP. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 17 / 50
  • 18. Boolean Coded Apertures An optimal ensemble of four 64 × 64 boolean coded apertures. Each spatial coordinate in the ensemble contains only one 1-value entry. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 18 / 50
  • 19. Bernoulli Coded Apertures An ensemble of four 64 × 64 Bernoulli coded apertures. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 19 / 50
  • 20. Performance of coded apertures Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 20 / 50
  • 21. Original datacube Boolean (40.4dB) Unsigned grayscale (31.2dB) Binary (27.7dB) Hadamard (27.7dB) Signed grayscale (22.7dB) Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 21 / 50
  • 22. Original datacube Boolean Unsigned grayscale Binary Hadamard Signed grayscale Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 22 / 50
  • 23. Coded Aperture Optimization: Spectral Selectivity UAV sensor requirements depend on flight duration, range, altitude, etc. Need: Hyperspectral imaging that dynamically adapts to optimal spectral bands. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 23 / 50
  • 24. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 24 / 50
  • 25. ((a)) Original ((b)) 12 Random Codes ((c)) 9 Optimal Codes The resulting spectral data cubes are shown as they would be viewed by a Stingray F-033C CCD Color Camera. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 25 / 50
  • 26. ((a)) Random Code ((b)) Original ((c)) Optimal Code ((d)) Random Codes ((e)) Optimized Codes Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 26 / 50
  • 27. ((a)) Random ((b)) Optimized ((c)) ((d)) Differences between the original and the reconstructed 3rd spectral channel (479nm) Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 27 / 50
  • 28. ((a)) Original ((b)) 12 Random Codes ((c)) 12 Optimized Codes The resulting spectral data cubes are shown as they would be viewed by a Stingray F-033C CCD Color Camera. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 28 / 50
  • 29. Coded Aperture Optimization: Image Classification Goal: classification of a spectral scene using Compressive measurements Optimal code aperture Sparsity signal model Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 29 / 50
  • 30. Given the FPA measurements g, every test pixel fi belongs to one of the P known classes {H(1) , H(2) , ..., H(P) } 0 10 20 30 40 50 60 70 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Spectrum Band Glutamine 0 10 20 30 40 50 60 70 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Spectrum Band Histidine 0 10 20 30 40 50 60 70 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Spectrum Band Isoleucine If a pixel fi ∈ H(k), its spectral profile lies in a low-dimensional subspace spanned by the training samples: {s (k) j }j=1,...,Np fi ≈ [s (k) 1 , s (k) 2 , ..., s (k) Nk ][α (k) 1 , α (k) 2 , ..., α (k) Np ]T = S(k) α(k) where, α(k) is a sparse vector.Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 30 / 50
  • 31. Sparsity Model Combining all class sub-dictionaries fi ≈ [S(1) , ..., S(k) , ..., S(P) ][α(1) , ..., α(k) , ..., α(P) ]T = Sα Ideally, if fi ∈ H(k), then α(j) = 0; ∀j = 1, ..., P; j = k. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 31 / 50
  • 32. Superposition of 3 shots for the different codes Optimal codes 5 10 15 20 25 30 5 10 15 20 25 30 0 0.5 1 1.5 2 2.5 3 Hadamard Codes 5 10 15 20 25 30 5 10 15 20 25 30 0 0.5 1 1.5 2 2.5 3 Bernoulli codes 5 10 15 20 25 30 5 10 15 20 25 30 0 0.5 1 1.5 2 2.5 3 Ck = K−1 k=0 (tk m,n)2 at each (m, n) spatial location. Optimal Hadamard Bernoulli Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 32 / 50
  • 33. Image Classification Results 200 spectral bands. 50 shots. Ground-truth Single spectral band SVM-full datacube (73.5%) Bernoulli-25%datacube (64.3%) Hadamard-25%datacube (63.4%) Optimal-25%datacube (73.7%) Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 33 / 50
  • 34. Image Classification Results 200 spectral bands. 50 shots. Ground-truth FPA measurement SVM-full datacube (73.5%) Bernoulli-25%datacube (64.3%) Hadamard-25%datacube (63.4%) Optimal-25%datacube (73.7%) Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 34 / 50
  • 35. Colored Coded Aperture Spectral Imaging Patterned coating combines micro-lithography with optical coating technology. Precision patterned coating and patterns Sub-pixel alignment accuracy Ultraviolet, visible, NIR, SWIR Multi-filter arrays on monolithic substrates Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 35 / 50
  • 36. NEW FAMILY OF CODED APERTURES Boolean Spectrally Selective Super-resolution Colored Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 36 / 50
  • 37. Colored coded aperture model Colored coded aperture is a color filter array Each entry is a wavelength selective color filter 3D Mask model has the same dimensions than the objective discrete data cube Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 37 / 50
  • 38. Linear dispersion and focal plane array integration Linear shifting operation Focal plane array (FPA) projections The number of pixels of the FPA detector is N(N + L − 1) N2 L (size of the spectral data cube) Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 38 / 50
  • 39. Random Boolean Code Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 39 / 50
  • 40. 4 Colors Random Code Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 40 / 50
  • 41. Restricted Isometry Property of Colored CASSI A = HΨΨΨ, fff = ΨΨΨθθθ, ΨΨΨ = W ⊗ ΨΨΨ2D Definition (1 − δs) ||θθθ||2 2 ≤ ||Aθθθ||2 2 ≤ (1 + δs) ||θθθ||2 2, δs = max T ⊂[N2L],|T |≤S ||AT |T |A|T | − I||2 2, A|T |, |T | columns of A indexed by the set T δs = max T ⊂[N2L],|T |≤S λmax A|T ||T | − I A|T | ir = hiψψψΩr = L−1 k=0 t i k mi −kN Ψmi +k(N ),Ωr (hi )j =    t i kj i− i V−kj N , if i − i V = j − kj N 0, otherwise, A = HΨΨΨ2D A = H W ⊗ ΨΨΨ2D Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 41 / 50
  • 42. A|T ||T | = AT |T |A|T | A|T ||T | r,u = K−1 =0 V−1 i=0 L−1 k1=0 L−1 k2=0 tk1 i−k1N tk2 i−k2N Ψi+k1N ,ΩrΨi+k2N ,Ωu Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 42 / 50
  • 43. Results: LH-Colored Coded Aperture A geometric interpretation of the colored coded apertures for LH-Colored filters (3 shots). Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 43 / 50
  • 44. Code Design Results B-Colored Coded Apertures (Band Pass Filters) Geometric interpretation of colored coded apertures for B-filters (3 shots) Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 44 / 50
  • 45. 461nm 470nm 479nm 488nm 497nm 506nm 515nm 524nm 533nm 542nm 551nm 560nm 569nm 578nm 587nm 596nm 16 channels 461-596nm 256 × 256 pixels Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 45 / 50
  • 46. Reconstruction From B-Colored Coded Apertures 2 4 6 8 10 12 14 30 40 50 60 70 80 90 Number of Shots PSNR Block Unblock Random B-colored B-colored Mean PSNR of the reconstructed data cubes. Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 46 / 50
  • 47. Reconstructions from Real Measurements (RGB) K = 1 K = 1 Sh=1 (i) 1 snapshot: Photomask, Coloring, Optimal K = 4 K = 4 Sh=4 (j) 4 snapshots: Photoask, Coloring, Optimal Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 47 / 50
  • 48. Reconstruction Results (Photomask vs Colored Coded Apertures) 454 nm K = 1 468 nm K = 1 485 nm K = 1 506 nm K = 1 529 nm K = 1 557 nm K = 1 594 nm K = 1 639 nm K = 1 454 nm K = 1 468 nm K = 1 485 nm K = 1 506 nm K = 1 529 nm K = 1 557 nm K = 1 594 nm K = 1 639 nm K = 1 (a) One snapshot reconstruction. (First row) Photomask, (Second row) Coloring 454 nm K = 4 468 nm K = 4 485 nm K = 4 506 nm K = 4 529 nm K = 4 557 nm K = 4 594 nm K = 4 639 nm K = 4 454 nm K = 4 468 nm K = 4 485 nm K = 4 506 nm K = 4 529 nm K = 4 557 nm K = 4 594 nm K = 4 639 nm K = 4 (b) 4 snapshots reconstructions. (First row) Photomask, (Second row) Coloring Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 48 / 50
  • 49. Spectral Reconstruction From Colored Coded Apertures Sh=4 P1 P2 454 468 485 506 529 557 594 639 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Nanometers Norm.Intensity Spectral Comparison: 4 Snapshots Spectrometer Photomask Coloring Optimal 454 468 485 506 529 557 594 639 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Nanometers Norm.Intensity Spectral Comparison: 4 Snapshots Spectrometer Photomask Coloring Optimal Pixel P1 Pixel P2 Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 49 / 50
  • 50. Conclusions Optical Coding for Compressive Spectral Imaging Good codes for reconstruction, classification, unmixing Colored codes offer multidimensional coding - open problem Convolution Optical Coding (Projections) Light field imaging X-ray tomography Super-resolution microscopy Gonzalo R. Arce () Compressive Spectral Image Sensing Oct., 2013 50 / 50