OPTIMED: Optimisation Iterative pour la Résolution de Problèmes Inverses de Grande Taille

OPTIMED: Optimisation Iterative pour la Résolution de Problèmes Inverses de Grande Taille

Lotfi Chaari Emilie Chouzenoux  Nelly Pustelnik  Caroline Chaux  Saïd Moussaoui 

INRIA Rhône-Alpes, Projet MISTIS

Université Paris-Est, IGM (UMR CNRS 8049)

Laboratoire de Physique, ENS Lyon (UMR CNRS 5672)

LUNAM Université, Ecole Centrale Nantes, IRCCyN (UMR CNRS 6597)

Page: 
329-374
|
DOI: 
https://doi.org/10.3166/TS.28.329-374
Received: 
N/A
| |
Accepted: 
N/A
| | Citation

OPEN ACCESS

Abstract: 

This paper deals with image reconstruction using iterative algorithms. The framework under consideration involves large amount of data and/or large observation operator. We proposed here algorithms for which we were able to state convergence proofs. The reliability and accuracy of the proposed approaches are demonstrated via three applications: the positron emission tomography (PET), the parallel magnetic resonance imaging (MRI) and twodimensional nuclear magnetic resonance (2D NMR).

Extended Abstract

The main objective of the ANR project OPTIMED (MDMSA program) was to develop new reconstruction methods to solve large-scale inverse problems arising in medical imaging. Two different modalities have been studied: the Positron Emission Tomography (PET) and the Magnetic Resonance Imaging (MRI). Furthermore, the application area has been extended to two dimensional Nuclear Magnetic Resonance (2D NMR) relaxometry. In these problems, existing convex optimization methods were unable to deal efficiently with the large amount of data and constraints, or to deliver reliable numerical solutions in reasonable time.

From a methodological point of view, the goal was to bring solution to large scale inverse problems (3D, 3D+t...) by solving optimization problems where the objective function accounts for the physical properties of the acquisition system as well as any prior information one can have on the imaged object. Two main avenues of research were therefore explored:

– Improve the reconstructed image quality by considering an adequate criterion in accordance with the noise statistical property (Poissonian or Gaussian) and by introducing relevant prior knowledge. Among all the possible priors, we favor the one promoting sparsity of the data after some linear transform (e.g. Wavelet frame decomposition), which appears to be a judicious choice when dealing with large amount of data (reduction of the memory load/swap time and introduction of regularity constraints in a more tractable fashion). Furthermore, the Lagrangian formulation of the constraints that have to be fulfilled by the solution, require the addition of a Barrier function in the criterion. This can lead to non-smooth optimization problems involving non-differentiable objective and/or constraint functions.

– Reduce the computational time that can be crucial for medical diagnosis. That is why, it was necessary to develop and implement iterative algorithms that had a convergence speed as high as possible. A compromise between convergence speed and iteration cost has thus to be made. Furthermore, in order to take into account the new available parallel architectures, we proposed splitting optimization algorithms well adapted to a parallelization process. In order to explore these avenues, two main families of optimization approaches have been explored during this study: proximal and majorization-minimization (MM) methods.

Proximal approaches

The proximal methods refer to iterative algorithms enable to minimize a sum of convex functions non-necessarily differentiable and that can involve linear perators. This class of algorithms is based on proximity operators that is a generalization of the projection onto a convex set. The relevance of these algorithms is mainly due to their strong convergence properties, their robustness to numerical errors, and their good numerical behavior. In the recent literature related to signal and image restoration, numerous works have considered proximal algorithms in order to efficiently deal with 1 -prior, wavelet-frame representations, non-necessarily Lipschitz differentiable data fidelity term, or additional constraints such as the data positivity.

Majorization-Minimization

At each iteration of a MM algorithm, a tangent majorant function, that majorizes the objective function and is equal to it at the current iterate, is constructed. The next iterate is obtained by minimizing this tangent majorant function, resulting in a sequence of simple iterates that reduces the cost function monotonically. Recently, an efficient MM quadratic line search procedure for descent algorithms has been introduced. In this strategy, the stepsize value results from successive minimizations of quadratic tangent majorant functions for the restriction of the cost function along the search direction. This method benefits from strong convergence results. We prove that it generalizes into a multi-dimensional search that ensures the convergence of several subspace algorithms. Quadratic MM strategies exclude the case of barrier criteria. Such functions are encountered for example in maximum entropy reconstruction and Positron Emission Tomography. We propose an MM line search algorithm for barrier functions, based on a quadratic majorant function augmented with a logarithmic term. This leads to a simple line search that ensures the convergence of several classical descent optimization strategies.

In summary, the ANR project OPTIMED offered an opportunity to apply modern tools of convex analysis and the theory of monotonic operators to solve large scale optimization problems. All these developments have been efficiently applied to real problems on image reconstruction in PET, parallel MRI and 2D NMR relaxometry.

RÉSUMÉ

Cet article s’intéresse au développement d’algorithmes itératifs pour la reconstruction d’images. La particularité de ces problèmes réside dans la taille élevée des images et la quantité importante de données générée par le processus de mesure. Outre le fait de surmonter la difficulté liée à la grande taille du problème, les algorithmes proposés se fondent sur des résultats de convergence théoriquement établis. Nous illustrons ces développements à l’aide de trois modalités d’imagerie distinctes : la tomographie par émission de positrons (TEP), l’imagerie par résonance magnétiques nucléaire (IRM) parallèle et la résonance magnétique nucléaire bidimentionnelle (RMN 2D).

Keywords: 

convex optimization, iterative algorithms, PET, MRI, 2D-NMR.

MOTS-CLÉS

optimisation convexe, algorithmes itératifs, TEP, IRM, RMN-2D.

1. Introduction
2. Quelques Schémas Itératifs pour la Résolution du Problème d’Optimisation
3. Reconstruction d’Images en Tomographie par Emission de Positrons
4. Reconstruction Régularisée en IRM Parallèle
5. Reconstruction de Spectres en Relaxométrie RMN Bidimensionnelle
6. Conclusion
  References

Afonso M., Bioucas-Dias J., Figueiredo M. (2011a). « Hybrid synthesis-analysis frame-based regularization: a criterion and an algorithm », Signal processing with adaptive sparse structured representations - SPARS11, Edinburgh, Scotland, 27-30 Jun.

Afonso M. V., Bioucas-Dias J. M., Figueiredo M. A. T. (2011b). « An augmented Lagrangian approach to the constrained optimization formulation of imaging inverse problems », IEEE Trans. Image Process., vol. 20, n° 3, p. 681-695, Mar.

Bauschke H. H., Combettes P. L. (2011). Convex Analysis and Monotone Operator Theory in Hilbert Spaces, CMS Books in Mathematics, Springer.

Beck A., Teboulle M. (2009). « Gradient-based algorithms with applications to signal-recovery problems », in Y., D. Palomar (eds), Convex Optimization in Signal Processing and Communications, Cambridge University Press.

Bect J., Blanc-Féraud L., Aubert G., Chambolle A. (2004). « A 1 -unified variational framework for image restoration », in T. Pajdla, J. Matas (eds), Proc. European Conference on Computer Vision, vol. LNCS 3024, Springer, Prague, Czech Republic, p. 1-13, May.

Bertero M., Boccacci P., Desidera G., Vicidomini G. (2009). « Image deblurring with Poisson data : from cells to galaxies », IEEE Trans. Image Process.

Bertsekas D. P. (1999). Nonlinear Programming, 2ème edn, Athena Scientific, Belmont, MA, USA.

Böhning D., Lindsay B. G. (1988). « Monotonicity of quadratic-approximation algorithms », Ann. Inst. Statist. Math., vol. 40, n° 4, p. 641-663.

Bredies K., Lorenz D. A. (2008). « Linear convergence of iterative soft-thresholding », J. Fourier Anal. Appl., vol. 14, n° 5-6, p. 813-837, Dec.

Briceño-Arias L. M., Combettes P. L. (2011). « A Monotone + Skew Splitting Model for Composite Monotone Inclusions in Duality », SIAM J. Optim..

Butler J. P., Reeds J. A., Dawson S. V. (1981). « Estimating Solutions of First Kind Integral Equations with Nonnegative Constraints and Optimal Smoothing», SIAM J. Numer. Anal., vol. 18, n° 3, p. 381-397, juin.

Canet D., Boubel J.-C., Canet-Soulas E. (2002). La RMN 2-ième édition : concepts, méthodes et applications, 2ème edn, Dunod, Paris.

Chaari L., Mériaux S., Badillo S., Pesquet J.-C., Ciuciu P. (2011a). « Multidimensional Wavelet-based Regularized Reconstruction for Parallel Acquisition in neuroimaging », submitted to IEEE Trans. on Medical Imaging.

Chaari L., Pesquet J.-C., Benazza-Benyahia A., Ciuciu P. (2008). « Autocalibrated Parallel MRI Reconstruction in the Wavelet Domain », Proc. IEEE ISBI, Paris, France, p. 756-759, May 14-17.

Chaari L., Pesquet J.-C., Benazza-Benyahia A., Ciuciu P. (2011b). « A wavelet-based regularized reconstruction algorithm for SENSE parallel MRI with applications to neuroimaging », Medical Image Analysis, vol. 15, n° 2, p. 185-201, Nov.

Chambolle A. (2004). « An algorithm for total variation minimization and applications », J. Math. Imaging Vision, vol. 20, n° 1-2, p. 89-97, Jan.

Chambolle A., Pock T. (2011). « A first-order primal-dual algorithm for convex problems with applications to imaging », J. Math. Imaging Vision, vol. 40, n°1, p. 120-145.

Charbonnier P., Blanc-Féraud L., Aubert G., Barlaud M. (1997). « Deterministic edgepreserving reg-ularization in computed imaging », IEEE Trans. Image Process., vol. 6, p. 298-311.

Chaux C., Combettes P. L., Pesquet J.-C., Wajs V. R. (2007). « A variational formulation for frame-based inverse problems », Inverse Problems, vol. 23, n°4, p. 1495-1518, Aug.

Chaux C., Pesquet J.-C., Pustelnik N. (2009). « Nested Iterative algorithms for convex constrained image recovery problems », SIAM J. Imaging Sci., vol. 2,n °2, p. 730-762, June.

Chen G. H.-G., Rockafellar R. T. (1997). « Convergence rates in forward-backward splitting », SIAM J. Optim., vol. 7, p. 421-444.

Chen G., Teboulle M. (1994). « A proximal-based decomposition method for convex minimization problems », Math. Programm., vol. 64, p. 81-101.

Chen X., Sun J. (2002). « Global Convergence of Two-parameter Family of Conjugate Gradient Methods without Line Search », J. Comput. Appl. Math., vol. 146, p. 37-45.

Chiang Y.-W., Borbat P. P., Freed J. H. (2005). « Maximum entropy : A complement to Tikhonov regularization for determination of pair distance distributions by pulsed ESR », J. Magn. Reson., vol. 177, n° 2, p. 184-196, déc.

Chouzenoux E., Idier J., Moussaoui S. (2011a). « A Majorize-Minimize Strategy for Subspace Optimization Applied to Image Restoration », IEEE Trans. Image Process., vol. 20, n° 18, p. 1517-1528, juin.

Chouzenoux E., Moussaoui S., Idier J. (2009a). A Majorize-Minimize line search algorithm for barrier functions, Technical report, IRCCyN, nov. http://hal.archives-ouvertes.fr/hal-00362304.

Chouzenoux E., Moussaoui S., Idier J., Mariette F. (2009b). « Reconstruction d’un spectre RMN 2D par maximum d’entropie », Actes 22e coll. GRETSI, Dijon, sep.

Chouzenoux E., Moussaoui S., Idier J., Mariette F. (2010). « Efficient Maximum Entropy Reconstruction of Nuclear Magnetic Resonance T1-T2 Spectra », IEEE Trans. Signal Process., vol. 58, n° 12, p. 6040-6051, déc.

Chouzenoux E., Pesquet J.-C., Talbot H., Jezierska A. (2011b). « A Memory Gradient Algorithm for l2-l0 Regularization with Applications to Image Restoration», Proc. IEEE ICIP, Brussels-Belgique, sep.

Combettes P. L. (2009). « Iterative construction of the resolvent of a sum of maximal monotone operators », J. Convex Anal., vol. 16, n° 4, p. 727-748.

Combettes P. L., Dinh D˜ung, V˜u B. C. (2011). « Proximity for sums of composite functions », J. Math. Anal. Appl., vol. 380, n° 2, p. 680-688, Aug.

Combettes P. L., Pesquet J.-C. (2007a). « A Douglas-Rachford Splitting Approach to Nonsmooth Con-vex Variational Signal Recovery », IEEE J. Selected Topics Signal Process., vol. 1, n° 4, p. 564-574, Dec.

Combettes P. L., Pesquet J.-C. (2007b). « Proximal thresholding algorithm for minimization over orthonormal bases », SIAM J. Optim., vol. 18, n° 4, p. 1351-1376, Nov.

Combettes P. L., Pesquet J.-C. (2008). « A proximal decomposition method for solving convex variational inverse problems », Inverse Problems, vol. 24, n°6, p. x+27, Dec.

Combettes P. L., Pesquet J.-C. (2010). « Proximal splitting methods in signal processing », in H. H. Bauschke, R. Burachik, P. L. Combettes, V. Elser, D. R. Luke, H. Wolkowicz (eds), Fixed-Point Algorithms for Inverse Problems in Science and Engineering, Springer-Verlag, New York.

Combettes P. L., Wajs V. R. (2005). « Signal Recovery by proximal forward-backward splitting», Multiscale Model. and Simul., vol. 4, n° 4, p. 1168-1200, Nov.

Conn A. R., Gould N., Sartenaer A., Toint P. L. (1994). On iterated-subspace minimization methods for nonlinear optimization, Technical Report n° 94-069, Rutherford Appleton Laboratory, Oxfordshire, UK, mai. ftp://130.246.8.32/pub/reports/cgstRAL94069.ps.Z.

Daubechies I., Defrise M., De Mol C. (2004). « An iterative thresholding algorithm for linear inverse problems with a sparsity constraint », Comm. Pure Applied Math., vol. 57, n° 11, p. 1413-1457, Nov.

De Pierro A. R. (1995). « A modified expectation maximization algorithm for penalized likelihood estimation in emission tomography », IEEE Trans. Medical Imaging, vol. 14, n° 1, p. 132-137, mars.

Dembo R. S., Eisenstat S. C., Steihaug S. (1982). « Inexact Newton Methods », SIAM J. Numer. Anal., vol. 19, n° 2, p. 400-408, avr.

Dempster A. P., Laird N. M., Rubin D. B. (1977). « Maximum likelihood from incomplete data via the EM algorithm », J. R. Statist. Soc. B, vol. 39, p. 1-38.

Eggermont P. P. B. (1993). « Maximum entropy regularization for Fredholm integral equations of the first kind », SIAM J. Math. Anal., vol. 24, n° 6, p.1557-1576.

English A. E., Whittall K. P., Joy M. L. G., Henkelman R. M. (1991). « Quantitative twodimensional time correlation relaxometry », Magn. Reson. Med., vol.22, p. 425-434.

Erdogan H., Fessler J. A. (1999). « Monotonic Algorithms for Transmission Tomography », IEEE Trans. Medical Imaging, vol. 18, n° 9, p. 801-814, sep.

Esser E., Zhang X., Chan T. (2010). « A General Framework for a Class of First Order Primal-Dual Algorithms for Convex Optimization in Imaging Science », SIAM J. Imaging Sci., vol. 3, n° 4, p. 1015-1046.

Fessler J. A., Booth S. D. (1999). « Conjugate-Gradient Preconditiong Methods for Shift-Variant PET Image Reconstruction », IEEE Trans. Image Process., vol. 8, n° 5, p. 688-699,mai.

Fessler J. A., Erdogan H. (1998). « A paraboloidal surrogates algorithm for convergent penalized-likelihood emission image reconstruction », IEEE Nuclear Science Symposium, vol. 2, p. 1132-1135.

Fiacco A. V., McCormick G. P. (1967). « The sequential unconstrained minimization technique (SUMT) without parameters », Operations Research, vol. 15, n° 5,, p. 820-827, sep.

Figueiredo M. A. T., Nowak R. D. (2003). « An EM algorithm for wavelet-based image restoration », IEEE Trans. Image Process., vol. 12, n° 8, p. 906-916, Aug.

Friston K. J., Holmes A. P., Worsley K. J., Poline J. P., Frith C. D., Frackowiak R. S. J. (1994). « Statistical parametric map in functional imaging : A general linear approach », Human Brain Mapping, vol. 2, n° 4, p. 189-210, Apr.

Geman D., Reynolds G. (1992). « Constrained restoration and the recovery of discontinuities », IEEE Trans. Pattern Anal. Mach. Intell., vol. 14, n° 3, p. 367-383, mars.

Geman D., Yang C. (1995). « Nonlinear image recovery with half-quadratic regularization », IEEE Trans. Image Process., vol. 4, n° 7, p. 932-946, juil.

Gjedde A. (1982). « Calculation of cerebral glucose phosphorylation from brain uptake of glucose analogs in vivo : a re-examination », Brain Res., vol. 257, n° 2, p. 237-274.

Griswold M. A., Jakob P. M., Heidemann R. M., Nittka M., Jellus V., Wang J., Kiefer B., Haase A. (2002). « Generalized autocalibrating partially parallel acquisitions GRAPPA », Magnetic Resonance in Medicine, vol. 47, n° 6, p. 1202-1210, Jun.

Hansen P. C. (1998). Rank-deficient and discrete ill-posed problems : numerical aspects of linear inversion, SIAM, Philadelphia, PA, USA.

Hunter D. R., Kenneth L. (2004). « A Tutorial on MM Algorithms », Amer. Statist., vol. 58, n° 1, p. 30-37, fév.

Hürlimann M. D., Venkataramanan L. (2002). « Quantitative Measurement of Two-Dimensional Distribution Functions of Diffusion and Relaxation in Grossly Inhomogeneous Fields », J. Magn. Reson., vol. 157, p. 31-42.

Idier J. (2001). « Convex Half-Quadratic Criteria and Interacting Auxiliary Variables for Image Restoration », IEEE Trans. Image Process., vol. 10, n° 7, p.1001-1009, juil.

Jacobson M. W., Fessler J. A. (2007). « An Expanded Theoretical Treatment of Iteration-Dependent Majorize-Minimize Algorithms », IEEE Trans. Image Process., vol. 16, n° 10,p. 2411-2422, oct.

Labat C., Idier J. (2007). Convergence of truncated half-quadratic and Newton algorithms, with application to image restoration, Technical report, IRCCyN, juin. http://www.irccyn.ec-nantes.fr/~idier/pub/labat07b.pdf.

Labat C., Idier J. (2008). « Convergence of conjugate gradient methods with a closed-form stepsize formula », J. Optim. Theory Appl., vol. 136, n° 1, p. 43-60, jan.

Lamanna R. (2005). « On the inversion of multicomponent NMR relaxation and diffusion decays in heterogeneous systems », Concepts Magn. Reson. Part A, vol. 26A, n°2, p. 78-90.

Lange K., Fessler J. A. (1995). « Globally convergent algorithms for maximum a posteriori transmission tomography », IEEE Trans. Image Process., vol. 4, n°10, p. 1430-1438, oct.

Laue E. D., Skilling J., Staunton J., Sibisi S., Brereton R. G. (1985). « Maximum entropy method in nuclear magnetic resonance spectroscopy », J. Magn. Reson., vol. 62, n° 3, p. 437-452.

Liu B., Ying L., Steckner M., Jun X., Sheng J. (2007). « Regularized SENSE Reconstruction using iteratively refined total variation method », IEEE International Symposium on Biomedical Imaging (ISBI), p. 121-124, Washington DC, Apr. 12-15.

Mariette F. (2004). « Relaxation RMN et IRM : un couplage indispensable pour l’étude des produits alimentaires », C. R. Acad. Sci., vol. Chimie 7, p. 221-232.

Mariette F., Guillement J. P., Tellier C., Marchal P. (1996). « Continuous relaxation time distribution decomposition by MEM », Signal Treat. and SignaL Anal. in NMR, vol. 10, p. 218-234.

Miele A., Cantrell J. W. (1969). « Study on a memory gradient method for the minimization of functions », J. Optim. Theory Appl., vol. 3, n° 6, p. 459-470.

Mohammad-Djafari A., Demoment G. (1988). « Utilisation de l’Entropie dans les problèmes de restau-ration et de reconstruction d’images », IEEE Trans. Signal Process., vol. 5, n° 4, p. 235-248.

Moreau J. J. (1962). « Fonctions convexes duales et points proximaux dans un espace hilbertien », C. R. Acad. Sci., vol. 255, p. 2897-2899.

Moreau J. J. (1965). « Proximité et dualité dans un espace hilbertien », Bull. Soc. Math. France, vol. 93, p. 273-299.

Murray W., Wright M. H. (1994). « Line Search Procedures for the Logarithmic Barrier Function », SIAM J. Optim., vol. 4, n° 2, p. 229-246.

Narkiss G., Zibulevsky M. (2005). Sequential Subspace Optimization Method for Large-Scale Unconstrained Problems, Technical Report n° 559, Israel Institute of Technology, oct.

Nash S. G. (2000). « A survey of truncated-Newton methods », J. Comput. Appl. Math., vol. 124, p. 45-59.

Ogawa S., Lee T. M., Kay A. R., Tank D. W. (1990). « Brain magnetic resonance imaging with contrast dependent on blood oxygenation », Proceedings of the National Academy of Sciences of the United States of America, vol. 87, p. 9868-9872, Dec.

Ortega J. M., Rheinboldt W. C. (1970). Iterative solution of nonlinear equations in several variables, Academic Press, New York, USA.

Patlak C. S., Blasberg R. G., Fenstermacher J. D. (1983). « Graphical evaluation of blood-tobrain transfer constants from multiple-time uptake data », J. Cereb. Blood Flow Metab., vol. 3, n° 1, p. 1-7, Mar.

Pesquet J.-C., Pustelnik N. (2011). « A parallel inertial proximal optimization method ». http ://www.optimization-online.org/DB_HTML/2010/11/2825.html.

Pichon C., Thiébaut E. (1998). « Non-Parametric recontruction of distribution functions from observed galactic discs », Month. Not. Roy. Astr. Soc., vol. 301, n° 2, p. 419-434.

Pinel P., Thirion B., Mériaux S., Jobert A., Serres J.and Le Bihan D., Poline J.-B., Dehaene S. (2007). « Fast reproducible identification and large-scale databasing of individual functional cognitive networks », BMC Neurosci., vol. 8, n° 1, p. 91, Oct.

Press W. H., Teukolsky S. A., Vetterling W. T., Flannery B. P. (1992). Numerical Recipes : The Art of Scientific Computing, 3ème edn, Cambridge Univ. Press, New York.

Pruessmann K. P., Weiger M., Scheidegger M. B., Boesiger P. (1999). « SENSE : sensitivity encoding for fast MRI », Magnetic Resonance in Medicine, vol. 42, n° 5, p. 952-962, Jul.

Pustelnik N., Chaux C., Pesquet J.-C. (2011). « Parallel Proximal Algorithm for Image Restoration Using Hybrid Regularization », IEEE Trans. Image Process., vol. 20, n° 9, p. 2450-2462, Sep.

Pustelnik N., Chaux C., Pesquet J.-C., Sureau F. C., Dusch E., C C. (2009). « Adapted Convex Optimization Algorithm for Wavelet-Based Dynamic PET

Reconstruction », International Meeting on Fully Three-Dimensional Image Reconstruction in Radiology and Nuclear Medicine (Fully3D), Beijing, Chine, 5-10 Sep.

Rivera M., Marroquin J. (2003). « Efficient half-quadratic regularization with granularity control », Image and Vision Computing, vol. 21, n° 4, p. 345-357, avr.

Rudin L., Osher S., Fatemi E. (1992). « Nonlinear total variation based noise removal algorithms », Physica D, vol. 60, n° 1-4, p. 259-268, Nov.

Setzer S., Steidl G., Teuber T. (2010). « Deblurring Poissonian images by split Bregman techniques », J. Visual Communication and Image Representation, vol. 21, n° 3, p. 193-199.

Skilling J., Bryan R. K. (1984). « Maximum entropy image reconstruction : General algorithm », Month. Not. Roy. Astr. Soc., vol. 211, p. 111-124.

Song Y. Q., Venkataramanan L., Hürlimann M. D., Flaum M., Frulla P., Straley C. (2002). « T1-T2 correlation spectra obtained using a fast two-dimensional Laplace inversion », J. Magn. Reson., vol. 154, p. 261-268.

Sotthivirat S., Fessler J. A. (2002). « Image recovery using partitioned-separable paraboloidal surrogate coordinate ascent algorithms », IEEE Trans. Image Process., vol. 11, n° 3, p. 306-317, mars.

Starck J.-L., Murtagh F., Fadili M. J. (2010). Sparse Image and Signal Processing : Wavelets, Curvelets, Morphological Diversity, Cambridge University Press.

Sternin E. (2008). « Use of Inverse Theory Algorithms in the Analysis of Biomembrane NMR Data », Methods in Membrane Lipids, vol. 400, Humana Press, p. 103-125.

Sun J., Zhang J. (2001). « Global Convergence of Conjugate Gradient Methods without Line Search », Ann. Oper. Res., vol. 103, p. 161-173, mars.

Sureau F. C., Pesquet J.-C., Chaux C., Pustelnik N., Reader A. J., Comtat C., Trebossen R. (2008). « Temporal wavelet denoising of PET sinograms and images», Proc. Nuclear Science Symposium Conference Record, Dresden, Germany, 19-25 Oct.

Svaiter B. F. (2011). « Weak convergence on Douglas-Rachford method », SIAM Journal on Control and Optimization, vol. 49, n° 1, p. 280-287.

Tikhonov A. (1963). « Tikhonov Regularization of incorrectly posed problems », Soviet Mathematics Doklady, vol. 4, p. 1624-1627.

Valk P., Bailey D. L., Townsend D. W., Maisey M. N. (2004). Positron Emission Tomography. Basic science and clinical pratice, Springer.

Van Loan C. F. (2000). « The ubiquitous Kronecker product », J. Comput. Appl. Math., vol. 123, p. 85-100.

Venkataramanan L., Song Y., Hürlimann M. D. (2002). « Solving Fredholm integrals of the first kind with tensor product structure in 2 and 2.5 dimensions »,, IEEE Trans. Signal Process., vol. 50, n° 5, p. 1017-1026.

Verhaeghe J., Van De Ville D., Khalidov I., D’Asseler Y., Lemahieu I., Unser M. (2008). « Dynamic PET Reconstruction Using Wavelet Regularization with Adapted Basis Functions », IEEE Trans. Med. Imag., vol. 27, n° 7, p. 943-959, July.

Wright M. H. (1995). « Why a pure primal Newton barrier step may be infeasible », SIAM J. Optim., vol. 5, p. 1-12.

Yuan Y. (2007). « Subspace techniques for nonlinear optimization », Some Topics in Industrial and Applied Mathematics, vol. 8, Series on Concrete and Applicable Mathematics, p. 206-218.

Zheng J., Saquib S. S., Sauer K., Bouman C. A. (2004). « Parallelizable Bayesian Tomography Algorithms with rapid, guarenteed convergence », SIAM J. Optim, vol. 14, n° 4, p. 1043-1056.