. A Student-t based sparsity enforcing hierarchical prior for linear inverse problems and its efficient Bayesian computation for 2D and 3D Computed Tomography Ali Mohammad-Djafari, Li Wang, Nicolas Gac and Folker Bleichrodt Laboratoire des Signaux et Syst` emes (L2S) UMR8506 CNRS-CentraleSup´ elec-UNIV PARIS SUD SUPELEC, 91192 Gif-sur-Yvette, France http://lss.centralesupelec.fr Email: [email protected]http://djafari.free.fr http://publicationslist.org/djafari iTwist2016, Aug. 24-26, 2016, Aalborg, Denemark A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 1
27
Embed
A Student-t based sparsity enforcing hierarchical prior ...djafari.free.fr/2016/pdf/iTwist2016_slides.pdf · Projections I Multi-Resolution implementation 7.Some results 8.Conclusions
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
.
A Student-t based sparsity enforcing hierarchicalprior for linear inverse problems and its efficientBayesian computation for 2D and 3D Computed
Tomography
Ali Mohammad-Djafari, Li Wang, Nicolas Gacand
Folker BleichrodtLaboratoire des Signaux et Systemes (L2S)
UMR8506 CNRS-CentraleSupelec-UNIV PARIS SUDSUPELEC, 91192 Gif-sur-Yvette, France
A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 1/27
Contents
1. Computed Tomography in 2D and 3D
2. Main classical methods
3. Basic Bayesian approach
4. Sparsity enforcing models through Student-t and IGSM
5. Computational tools: JMAP, EM, VBA
6. Implementation issuesI Main GPU implementation steps: Forward and Back
ProjectionsI Multi-Resolution implementation
7. Some results
8. Conclusions
A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 2/27
Computed Tomography: Seeing inside of a body
I f (x , y) a section of a real 3D body f (x , y , z)
I gφ(r) a line of observed radiography gφ(r , z)
I Forward model:Line integrals or Radon Transform
gφ(r) =
∫Lr,φ
f (x , y) dl + εφ(r)
=
∫∫f (x , y) δ(r − x cosφ− y sinφ) dx dy + εφ(r)
I Inverse problem: Image reconstruction
Given the forward model H (Radon Transform) anda set of data gφi (r), i = 1, · · · ,Mfind f (x , y)
A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 3/27
2D and 3D Computed Tomography
3D 2D
gφ(r1, r2) =
∫Lr1,r2,φ
f (x , y , z) dl gφ(r) =
∫Lr,φ
f (x , y) dl
Forward probelm: f (x , y) or f (x , y , z) −→ gφ(r) or gφ(r1, r2)Inverse problem: gφ(r) or gφ(r1, r2) −→ f (x , y) or f (x , y , z)
A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 4/27
Algebraic methods: Discretization
f (x , y)
-x
6y
����� @
@@
���@@
HHH
���������������r
φ
•D
g(r , φ)
S•
@@
@@
@@@
@@
@@@
�
�
��
�������
fN
f1
fj
gi
HijQQQQQQQQ
QQ
f (x , y) =∑
j fj bj(x , y)
bj(x , y) =
{1 if (x , y) ∈ pixel j0 else
g(r , φ) =
∫Lf (x , y) dl gi =
N∑j=1
Hij fj + εi
gk = Hk f + εk , k = 1, · · · ,K −→ g = Hf + ε
gk projection at angle φk , g all the projections.A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 5/27
Algebraic methods
g =
g1...
gK
,H =
H1...
HK
→ gk = Hk f+εk → g =∑k
Hk f+εk = Hf+ε
I H is huge dimensional: 2D: 106 × 106, 3D: 109 × 109.I Hf corresponds to forward projectionI Htg corresponds to Back projection (BP)I H may not be invertible and even not squareI H is, in general, ill-conditionedI In limited angle tomography H is under determined, si the
problem has infinite number of solutionsI Minimum Norm Solution
f = Ht(HHt)−1g =∑k
Htk(HkHt
k)−1gk
can be interpreted as the Filtered Back Projection solution.A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 6/27
Prior information or constraints
I Positivity: fj > 0 or fj ∈ IR+
I Boundedness: 1 > fj) > 0 or fj ∈ [0, 1]
I Smoothness: fj depends on the neighborhoods.
I Sparsity: many fj are zeros.
I Sparsity in a transform domain: f = Dz and many zj arezeros.
I Discrete valued (DV): fj ∈ {0, 1, ...,K}I Binary valued (BV): fj ∈ {0, 1}I Compactness: f (r) is non zero in one or few
non-overlapping compact regions
I Combination of the above mentioned constraintsI Main mathematical questions:
I Which combination results to unique solution ?I How to apply them ?
A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 7/27
and ∇J (·) is the gradient of J (·).A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 14/27
Implementation issues
I In almost all the algorithms, the step of computation of fneeds an optimization algorithm. The criterion to optimize isoften in the form ofJ(f) = ‖g −Hf‖22 + λ‖Df‖22 or J(z) = ‖g −HDz‖2 + λ‖z‖1
I Very often, we use the gradient based algorithms which needto compute ∇J(f) = −2Ht(g −Hf) + 2λDtDf and so, thesimplest case, in each step, we have
f(k+1)
= f(k)
+ α(k)[Ht(g −Hf
(k)) + 2λDtDf
(k)]
1. Compute g = Hf (Forward projection)
2. Compute δg = g − g (Error or residual)
3. Compute δf1 = H′δg (Backprojection of error)
4. Compute δf2 = −D′Df and update f(k+1)
= f(k)
+ [δf1 + δf2]
I Steps 1 and 3 need great computational cost and have beenimplemented on GPU. In this work, we used ASTRA Toolbox.
A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 15/27
Multi-Resolution Implementation
Sacle 1: black g(1) = H(1)f(1) ( N × N )
Sacle 2: green g(2) = H(2)f(2) (N/2× N/2)
Sacle 3: red g(3) = H(3)f(3) (N/4× N/4)
A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 16/27
ResultsPhantom:128 x 128 x 128Projections:128, 64 and 32SNR=40 dB
128 projections 64 projections 32 projections
A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 17/27
ResultsPhantom:128 x 128 x 128Projections:128SNR=20 dB
QR TV HHBM
A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 18/27
Results with High SNR=30dB
δf = ‖f−f‖‖f‖ δg = ‖Hf−g‖
‖g‖
A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 19/27
Results: High SNR=30dB and Low SNR=20dB
δf = ‖f−f‖‖f‖ High SNR=30dB δf = ‖f−f‖
‖f‖ Low SNR=20dB
A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 20/27
Variational Bayesian Approximation (VBA)Depending on cases, we have to handlep(f,θ|g), p(f, z,θ|g), p(f,w,θ|g) or p(f,w, z,θ|g).
Let consider the simplest case:
I Approximate p(f,θ|g) by q(f,θ|g) = q1(f|g) q2(θ|g)and then continue computations.
The expressions of v zj , vεi and v ξ j are more complex and needs thecomputation of the diagonal elements of the posterior covariances.Specific techniques are needed to compute them efficiently.
v zj =βz0+
12<z2j >
αz0+3/2 , vεi =βε0+
12
(⟨g i−
[Hf]i
)2⟩αε0+3/2 , v ξ j =
βξ0+12
⟨(f j−[Dz]
j
)2⟩αξ0+3/2
A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 26/27
Conclusions
I Computed Tomography is an ill-posed Inverse problem
I Algebraic methods and Regularization methods push furtherthe limitations.
I Bayesian approach has many potentials for real applicationswhere we need to estimate the hyper parameters(semi-supervised) and to quantify the uncertainties.
I Hierarchical prior model with hidden variables are verypowerful tools for Bayesian approach to inverse problems.
I Main Bayesian computation tools: JMAP, VBA, AMP andMCMC
I Application in different other imaging systems: Microwaves,PET, Ultrasound, Optical Diffusion Tomography (ODT),..
I Current Projects: Efficient implementation in 3D cases toreconstruct 1024 x 1024 x 1024 volumes using GPU.
A. Mohammad-Djafari, Bayesian Discrete Tomography from a few number of projections, Mars 21-23, 2016, Polytechnico de Milan, Italy. 27/27