“Refractive Index Map Reconstruction in Optical ... · ELEN Wavelets and Sparsity XIV...

Preview:

Citation preview

ELEN Wavelets and Sparsity XIV

“Refractive Index Map Reconstruction in Optical Deflectometry

Using Total-Variation Regularization”

L. Jacques1, A. González2, E. Foumouo2 and P. Antoine2

1 Institute of Information and Communication Technologies, Electronics and Applied Mathematics (ICTEAM)

2 Institute of Condensed Matter and Nanosciences (IMCN)

University of Louvain, Louvain-la-Neuve, Belgium

ELEN Wavelets and Sparsity XIV

1. Introduction

ELEN Wavelets and Sparsity XIV

‣ World Before Sparsity: Human Readable Sensing (up to an orthogonal transform, e.g., Fourier)

Nyquist sampling (wrt bandlimitedness)

Examples: Photography, First MRIs, Interferometry, ...

Sparsity in Wavefields

3

World SensingDevice Human

Optimized

(1/2)

ELEN Wavelets and Sparsity XIV

‣ Paradigm shift allowed by Sparsity:Computer “Readable” Sensing + Prior information (sparsity)

Examples: Seismology, Radio-Interferometry, Compressed Sensing, MRI,Computed Photography, ... (Many in this Workshop!)

4

World SensingDevice Human

Sparsity in Wavefields

Optimized

(2/2)

ELEN Wavelets and Sparsity XIV

2. Optical Deflectometry

ELEN Wavelets and Sparsity XIV

Optical Computed Tomography Classification:‣ Absorption Tomography (X-ray tomography)

‣ Phase Tomography (using interferometry)

‣ and Deflection Tomography ...6

(Check Session 12, Tue 23)

ELEN Wavelets and Sparsity XIV

‣ Problem:Reconstructing the refractive index map of transparent materials from light deflection measurements under multiple orientations.

‣ Advantages of deflectometry:‣ Insensitive to vibrations (vs. interferometry)‣ Less sensitive to dispersive medium‣ Precision: up to 10nm flatness deviation on 50mm FOV

Interests: ‣ (transparent) object surface topology ‣ default detection in glass, crystal growth study

Deflectometric Framework:

7

multifocal lens

(e.g., on window glass)

ELEN Wavelets and Sparsity XIV

∆(τ, θ) ��

R2

�∇n(x) · pθ

�δ(τ − x · pθ) d

2x

Mathematical Model:

(first order approximation)

ELEN Wavelets and Sparsity XIV

Experimentally: Phase-Shifting Schlieren“Coding light deviation in intensity variations”

Incoherent Backlight Source

LCD based programmable spatial filter:

Telecentric Imaging Optics

Pinhole Spatial Filter

CCD camera

m(x) = sin( 2πΛ x)

Object Observation

Area

(1/2)

ELEN Wavelets and Sparsity XIV

∆x = f tanα

sin( 2πΛ ∆x)

∆x

Intensity change

Object

Phase-Shifting

Algorithm

m(x) = sin( 2πΛ x)

Experimentally: Phase-Shifting Schlieren(2/2)

ELEN Wavelets and Sparsity XIV

3. Reconstruction Methods

ELEN Wavelets and Sparsity XIV

Deflectometric Central Slice Theorem:

with �n the 2-D FFT of n.

∆(τ, θ) ��

R2

�∇n(x) · pθ

�δ(τ − x · pθ) d

2x

Sensing Model:

Continuous Facts:

z(ω, θ) :=

R∆(τ, θ) e−iτωdτ = iω �n

�ω pθ

⇒ z(ω, θ) = iω �n�ω pθ

ELEN Wavelets and Sparsity XIV

Refractive Index Prior‣ Heterogenous transparent materials

with slowly varying refractive indexseparated by sharp interfaces

13

(e.g., optical fibers)

The perfect “cartoon shape” model (BV, TV, ...)

“Sparse” gradient, i.e.,small Total Variation norm

�n�TV =

R�∇n(x)� dxn0

n1

ELEN Wavelets and Sparsity XIV

‣ Weighting:

‣ Polar to Cartesian interpolation:

Discretization of Measurements

14

y(ω, θ) :=z(ω, θ) + ε(ω, θ)

iω= �n

�ω pθ

�+ η(ω, θ).

Given z = Fτ (∆(·, θ))

with ε additive white Gaussian noise (in τ and ω)

(ω, θ) ∼ ωpθ → (kx, ky)

Controlling error: Cartesian grid resolution

(could be NFFT)

y(ω, θ) → �y(kx, ky)

kx

ky

(noise error)

(interpolation error)

ExperimentalCoordinateSystem

(1/2)

ELEN Wavelets and Sparsity XIV

‣ Recording interpolated frequency pixel locations:

‣ Final Discrete Inverse Problem: Tomography (but “colored”)

15

Discretization of Measurements

y = SF n+ η

K = {k1, · · · , kM/2} ⊂ R2

Selection Operator

2-D FT

noise (colored)interp. + meas.

y =

y(k1)

...y(kM/2)

∈ CM/2 � RM

∈ RN

Ill-posed: M � N

(2/2)

ELEN Wavelets and Sparsity XIV

Filtered Back Projection‣ In our notation: FBP amounts to solve

16

argminu∈RN

�u�2 s.t. y = SFu

⇒ n∗ = (SF )†y = F ∗S∗ y

since SS∗ = FF ∗ = Id

ELEN Wavelets and Sparsity XIV

Regularized Whitened Reconstruction‣ Using TV prior and whitening noise: TV-L2

‣ Estimating W : ‣ Montecarlo on Normal Gaussian‣ Estimating std. dev. from

‣ Solving the reconstruction: ‣ Proximal methods‣ Iterative Chambolle-Pock algorithm

17

argminu∈RN

�u�TV s.t. �W (y − SFu)�2 � �

with W = diag�σ(η1), · · · ,σ(ηM/2)

�−1

∆(τ, θ)

ELEN Wavelets and Sparsity XIV

4. Results

ELEN Wavelets and Sparsity XIV

Synthetic Images‣ Generating realistic maps:

‣ Using Image Reconstruction Toolbox (IRT1) for computing

‣ Adding noise and reconstructing FBP & TV-L2

19

1 http://www.eecs.umich.edu/~fessler/code/

∆(τ, θ)

τ

θ

Varying #θ, Fixed #τ

(1/4)

ELEN Wavelets and Sparsity XIV

‣ No meas. noise‣ M/N = 10%

20

Synthetic Images

4.94 dB 10.37 dB

(2/4)

ELEN Wavelets and Sparsity XIV

‣ Meas. noise: 11.51dB‣ M/N = 10%

21

Synthetic Images

4.34 dB 5.85 dB

(3/4)

ELEN Wavelets and Sparsity XIV

‣ Measurement noise = 20 dB (another image)

22

Synthetic Images

# angles

(4/4)

SNR

(dB

)

FBP

TV-L2

ELEN Wavelets and Sparsity XIV

Experimental Data‣ bundle of 10 fibers immersed

in an optical fluid‣ working on one z-slice (2-D problem)

‣ Origin of must be calibrated!‣ Noise is estimated on

23

#τ = 696

τ

∆(τ, θ)

(1/3)

ELEN Wavelets and Sparsity XIV

24

Experimental Data

FBP

#θ = 60,M/N � 30%

(2/3)

ELEN Wavelets and Sparsity XIV

25

Experimental Data

FBP TV-L2

#θ = 60,M/N � 30%

(2/3)

ELEN Wavelets and Sparsity XIV

Do not forget calibration!‣ If misaligned origin

FBP and TV-L2 will look something like this:

26

τ

(3/3)

ELEN Wavelets and Sparsity XIV

5. Conclusion

ELEN Wavelets and Sparsity XIV

Conclusion and Perspectives‣ Recent sparsity-driven reconstruction methods can be

applied to Optical Deflectometry.

28

ELEN Wavelets and Sparsity XIV

Conclusion and Perspectives‣ Recent sparsity-driven reconstruction methods can be

applied to Optical Deflectometry.‣ Outside of easy synthetic Cartezian worlds,

hard time with gridding/interpolation/noise/calibration/...

29

ELEN Wavelets and Sparsity XIV

Conclusion and Perspectives‣ Recent sparsity-driven reconstruction methods can be

applied to Optical Deflectometry.‣ Outside of easy synthetic Cartezian worlds,

hard time with gridding/interpolation/noise/calibration/... ‣ Same framework valid for:

Phase-Contrast X-Ray Tomography (& no first order approx!)

30

ELEN Wavelets and Sparsity XIV

Conclusion and Perspectives‣ Recent sparsity-driven reconstruction methods can be

applied to Optical Deflectometry.‣ Outside of easy synthetic Cartezian worlds,

hard time with gridding/interpolation/noise/calibration/... ‣ Same framework valid for:

Phase-Contrast X-Ray Tomography (& no first order approx!) ‣ Future Improvements:

‣ Integrating positivity constraint‣ Using NFFT and reduce interpolation noise‣ Testing other reconstruction (Dantzig Fidelity ?)

31

�Φ∗(y − Φn)�∞ � ρforces spatially stationary noise

ELEN Wavelets and Sparsity XIV

Thank you!

Recommended