Deep-plug-and-play proximal Gauss-Newton method with applications to nonlinear, ill-posed inverse problems. (English) Zbl 07725199
Summary: In this paper we propose a proximal Gauss-Newton method for the penalized nonlinear least squares optimization problem arising from regularization of ill-posed nonlinear inverse problems. By exploiting the modular structure that characterizes the proximal-type methods, we plug in a pre-trained graph neural net denoiser in place of the standard proximal map. This allows to mould the prior on the data. An encoder-decoder Graph U-Net architecture is proposed as denoiser, which works on unstructured data; its mathematical formulation is derived to analyse the Liptschitz condition. With the intent of showing the benefits of applying deep Plug-and-Play reconstructions, we consider as an exemplar application, the nonlinear Electrical Impedance Tomography, a promising non-invasive imaging technique mathematically formulated as a highly nonlinear ill-posed inverse problem.
MSC:
| 65K10 | Numerical optimization and variational techniques |
| 68T20 | Problem solving in the context of artificial intelligence (heuristics, search strategies, etc.) |
Keywords:
ill-posed inverse problems; proximal Gauss-Newton; nonlinear optimization; deep-PnP; graph neural networksReferences:
| [1] | S. J. P. Becker Fadili Ochs, On Quasi-Newton forward-backward splitting: Proximal calculus and convergence, SIAM Journal on Optimization, 29, 2445-2481 (2019) · Zbl 1461.65128 · doi:10.1137/18M1167152 |
| [2] | A. B. M. A. W. R. B. Borsic Graham Adler Lionheart, In vivo impedance imaging with total variation regularization, IEEE Transactions on Medical Imaging, 29, 44-54 (2010) · doi:10.1109/TMI.2009.2022540 |
| [3] | A. Buades, B. Coll and J.-M. Morel, A non-local algorithm for image denoising, in Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit., (2005), 60-65. |
| [4] | G. T. S. H. S. C. A. Buzzard Chan Sreehari Bouman, Plug-and-play unplugged: Optimization-free reconstruction using consensus equilibrium, SIAM Journal on Imaging Sciences, 11, 2001-2020 (2018) · Zbl 1419.94002 · doi:10.1137/17M1122451 |
| [5] | A.-P. Calderón, On an inverse boundary value problem, In: Seminar on Numerical Analysis and its Applications to Continuum Physics, (1980), 65-73. |
| [6] | M. D. J. C. S. J. Cheney Isaacson Newell Simske Goble, NOSER: An algorithm for solving the inverse conductivity problem, Int. J. Imaging Syst. Technol., 2, 66-75 (1990) · doi:10.1002/ima.1850020203 |
| [7] | F. Colibazzi, D. Lazzaro, S. Morigi and A. Samoré, Learning nonlinear electrical impedance tomography, Journal of Scientific Computing, 90 (2022), Paper No. 58, 23 pp. · Zbl 1487.65132 |
| [8] | P. L. V. R. Combettes Wajs, Signal recovery by proximal forward-backward splitting, Multiscale Modeling & Simulation, 4, 1168-1200 (2005) · Zbl 1179.94031 · doi:10.1137/050626090 |
| [9] | K. A. V. K. Dabov Foi Katkovnik Egiazarian, Image denoising by sparse 3-D transform-domain collaborative filtering, IEEE Trans. Image Process., 16, 2080-2095 (2007) · doi:10.1109/TIP.2007.901238 |
| [10] | M. Fey and J. E. Lenssen, Fast graph representation learning with PyTorch Geometric, In ICLR Workshop on Representation Learning on Graphs and Manifolds, 2019. |
| [11] | H. H. B. Y. Y. Fu Liu Han Bo Yang Hu, A proximal iteratively regularized Gauss-Newton method for nonlinear inverse problems, Journal of Inverse and Ill-posed Problems, 25, 341-356 (2017) · Zbl 1365.65241 · doi:10.1515/jiip-2015-0092 |
| [12] | R. G. C. D. K. N. Gavaskar Athalye Chaudhury, On plug-and-play regularization using linear denoisers, IEEE Transactions on Image Processing, 30, 4802-4813 (2021) · doi:10.1109/TIP.2021.3075092 |
| [13] | R. G. K. N. Gavaskar Chaudhury, Plug-and-play ista converges with kernel denoisers, IEEE Signal Processing Letters, 27, 610-614 (2020) · doi:10.1109/LSP.2020.2986643 |
| [14] | D. C. Geman Yang, Nonlinear image recovery with half-quadratic regularization, IEEE Transactions on Image Processing, 4, 932-946 (1995) · doi:10.1109/83.392335 |
| [15] | H. E. B. M. J. Gouk Frank Pfahringer Cree, Regularisation of neural networks by enforcing Lipschitz continuity, Machine Learning, 110, 393-416 (2021) · Zbl 07432806 · doi:10.1007/s10994-020-05929-w |
| [16] | S. J. A. Hamilton Hauptmann, Deep D-bar: Real-time electrical impedance tomography imaging with deep neural networks, IEEE Transactions on Medical Imaging, 37, 2367-2377 (2018) · doi:10.1109/TMI.2018.2828303 |
| [17] | K. Hammernik, Learning a variational network for reconstruction of accelerated MRI data, Magnetic Resonance in Medicine, 79, 3055-3071 (2018) · doi:10.1002/mrm.26977 |
| [18] | M. Hanke, A regularizing Levenberg - Marquardt scheme, with applications to inverse groundwater filtration problems, Inverse Problems, 13, 79-95 (1997) · Zbl 0873.65057 · doi:10.1088/0266-5611/13/1/007 |
| [19] | N. A. S. T. Hänninen Pulkkinen Arridge Tarvainen, Adaptive stochastic Gauss-Newton method with optical Monte Carlo for quantitative photoacoustic tomography, J. Biomed Opt., 27, 083013 (2022) · doi:10.1117/1.JBO.27.8.083013 |
| [20] | A. Hauptmann, V. Kolehmainen, N. M. Mach, T. Savolainen, A. Seppänen and S. Siltanen, 2D electrical impedance tomography dataset, https://doi.org/10.5281/zenodo.1203914, 2017. |
| [21] | A. Hauptmann, V. Kolehmainen, N. M. Mach, T. Savolainen, A. Seppänen and S. Siltanen, Open 2D Electrical Impedance Tomography data archive, https://doi.org/10.48550/arXiv.1704.011782017, 2017. |
| [22] | W. D. B. A. S. J. Herzberg Rowe Hauptmann Hamilton, Graph convolutional networks for model-based learning in nonlinear inverse problems, IEEE Transactions on Compuational Imaging, 7, 1341-1353 (2021) · doi:10.1109/TCI.2021.3132190 |
| [23] | G. S. Hongyang Ji, Graph U-Nets, Proceedings of the 36th International Conference on Machine Learning, 97, 2083-2092 (2019) |
| [24] | M. Huska, D. Lazzaro and S. Morigi, A forward-backward strategy for handling non-linearity in Electrical Impedence Tomography, Lecture Notes LNCS 12951 Part III, Computational Science and Its Applications, Eds: Gervasi, O., Murgante, B., Misra, S., Garau, C., Blečić, I., et al., ISBN 978-3-030-86970-0, 2021,635-651. |
| [25] | M. Huska, D. Lazzaro, S. Morigi, A. Samore and G. Scrivanti, Spatially-adaptive variational reconstructions for linear inverse electrical impedance tomography, J. Sci. Comput., 84 (2020), 29 pp. · Zbl 07246751 |
| [26] | J. P. A. T. Jauhiainen Kuusela Sepp \({{\rm{\ddot n}}}\) en Valkonen, Relaxed Gauss-Newton methods with applications to electrical impedance tomography, SIAM Journal on Imaging Sciences, 13, 1415-1445 (2020) · Zbl 1465.65090 · doi:10.1137/20M1321711 |
| [27] | F. Jia, W. H. Wong and T. Zeng, DDUNet: Dense dense U-net with applications in image denoising, 2021 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW), 2021,354-364. |
| [28] | U. S. H. B. Kamilov Mansour Wohlberg, A Plug-and-play priors approach for solving nonlinear imaging inverse problems, IEEE Signal Processing Letters, 24, 1872-1876 (2017) · doi:10.1109/LSP.2017.2763583 |
| [29] | T. N. Kipf and M. Welling, Semi-Supervised Classification with Graph Convolutional Networks, International Conference on Learning Representations, 2017. |
| [30] | A. Lanza, S. Morigi and F. Sgallari, Convex image denoising via non-convex regularization, In J.-F. Aujol, M. Nikolova, and N. Papadakis, editors, Proc. Scale Space and Variational Methods in Computer Vision (SSVM), Springer International Publishing, 2015,666-677. · Zbl 1444.94015 |
| [31] | J. D. Y. M. A. Lee Sun Saunders, Proximal Newton-type methods for minimizing composite functions, SIAM Journal on Optimization, 24, 1420-1443 (2014) · Zbl 1306.65213 · doi:10.1137/130921428 |
| [32] | Y. S. Li Osher, A new median formula with applications to PDE based denoising, Communications in Mathematical Sciences, 7, 741-753 (2009) · Zbl 1187.49020 · doi:10.4310/CMS.2009.v7.n3.a11 |
| [33] | Z. J. Li Wu, Learning deep cnn denoiser priors for depth image inpainting, Applied Sciences, 9, 1103 (2019) · doi:10.3390/app9061103 |
| [34] | W. R. B. Lionheart, EIT Reconstruction algorithms: Pitfalls, challenges and recent developments, Physiological Measurement, 25, 125 (2004) · doi:10.1088/0967-3334/25/1/021 |
| [35] | B. B. C. J. M. Z. F. X. X. F. Liu Yang Xu Xia Dai Ji You Dong Shi Fu, pyEIT: A python based framework for electrical impedance tomography, SoftwareX, 7, 304-308 (2018) · doi:10.1016/j.softx.2018.09.005 |
| [36] | T. Meinhardt, et al., Learning proximal operators: Using denoising networks for regularizing inverse imaging problems, Proceedings of the IEEE International Conference on Computer Vision, 2017. |
| [37] | N. S. Parikh Boyd, Proximal algorithms, Found. Trends Optim., 3, 127-239 (2014) · doi:10.1561/2400000003 |
| [38] | Y. Park, S. Dhar, S. Boyd and S. Mohak, Variable metric proximal gradient method with diagonal Barzilai-Borwein stepsize, ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), (2020), 3597-3601. |
| [39] | A. Paszke, S. Gross, F. Massa, et al., Pytorch: An imperative style, high-performance deep learning library, In Advances in Neural Information Processing Systems (NIPS), (2019), 8024-8035. |
| [40] | J. V. M. J. E. P. Portilla Strela Wainwright Simoncelli, Image denoising using scale mixtures of gaussians in the wavelet domain, IEEE Transactions on Image Processing, 12, 1338-1351 (2003) · Zbl 1279.94028 · doi:10.1109/TIP.2003.818640 |
| [41] | R. G. Ramlau Teschke, Tikhonov replacement functionals for iteratively solving nonlinear operator equations, Inverse Problems, 21, 1571-1592 (2005) · Zbl 1078.47030 · doi:10.1088/0266-5611/21/5/005 |
| [42] | Y. M. P. Romano Elad Milanfar, The little engine that could: Regularization by denoising (RED), SIAM J. Imaging Sci., 10, 1804-1844 (2017) · Zbl 1401.62101 · doi:10.1137/16M1102884 |
| [43] | E. J. S. X. Z. W. Ryu Liu Wang Chen Wang Yin, Plug-and-play methods provably converge with properly trained denoisers, Proc. Int. Conf. Mach. Learn., 97, 5546-5557 (2019) |
| [44] | S. S. Salzo Villa, Convergence analysis of a proximal Gauss-Newton method, Computational Optimization and Applications, 53, 557-589 (2012) · Zbl 1282.90236 · doi:10.1007/s10589-012-9476-9 |
| [45] | A. M. D. P. L. S. Samoré Rusci Lazzaro Melpignano Benini Morigi, BrightNet: A deep CNN for OLED-Based point of care immunofluorescent diagnostic systems, IEEE Transactions on Instrumentation and Measurement, 69, 6766-6775 (2020) · doi:10.1109/TIM.2020.2973913 |
| [46] | I. A. S. F. Selesnick Lanza Morigi Sgallari, Non-convex total variation regularization for convex denoising of signals, J. Math. Imaging and Vision, 62, 825-841 (2020) · Zbl 1485.94026 · doi:10.1007/s10851-019-00937-5 |
| [47] | E. M. D. Somersalo Cheney Isaacson, Existence and Uniqueness for Electrode Models for Electric Current Computed Tomography, SIAM Journal on Applied Mathematics, 52, 1023-1040 (1992) · Zbl 0759.35055 · doi:10.1137/0152060 |
| [48] | S. S. V. B. G. T. L. F. J. P. C. A. Sreehari Venkatakrishnan Wohlberg Buzzard Drummy Simmons Bouman, Plug-and-play priors for bright field electron tomography and sparse interpolation, IEEE Trans. Comput. Imaging, 2, 408-423 (2016) |
| [49] | J. Sun, H. Li and Z. Xu, Deep ADMM-net for compressive sensing MRI, Advances in Neural Information Processing Systems, 29 (2016). |
| [50] | Y. Sun, B. Wohlberg and U. S. Kamilov, An online plug-and-play algorithm for regularized image reconstruction, IEEE Trans. Comput. Imaging, 5 (2019), 395-408. |
| [51] | Y. Z. X. B. U. S. Sun Wu Xu Wohlberg Kamilov, Scalable plug-and-play ADMM with convergence guarantees, IEEE Transactions on Computational Imaging, 7, 849-863 (2021) · doi:10.1109/TCI.2021.3094062 |
| [52] | T. M. S. R. Tarvainen Vauhkonen Arridge, Gauss-Newton reconstruction method for optical tomography using the finite element solution of the radiative transfer equation, Journal of Quantitative Spectroscopy and Radiative Transfer, 109, 2767-2778 (2008) · doi:10.1016/j.jqsrt.2008.08.006 |
| [53] | T. R. Tirer Giryes, Image restoration by iterative denoising and backward projections, IEEE Trans. Image Process., 28, 1220-1234 (2019) · Zbl 1409.94590 · doi:10.1109/TIP.2018.2875569 |
| [54] | T. R. Tirer Giryes, Super-resolution via image-adapted denoising cnns: Incorporating external and internal learning, IEEE Signal Processing Letters, 26, 1080-1084 (2019) · doi:10.1109/LSP.2019.2920250 |
| [55] | M. D. P. A. E. J. P. Vauhkonen Vadasz Karjalainen Somersalo Kaipio, Tikhonov regularization and prior information in electrical impedance tomography, IEEE Transactions on Medical Imaging, 17, 285-293 (1998) · doi:10.1109/42.700740 |
| [56] | S. V. Venkatakrishnan, C. A. Bouman and B. Wohlberg, Plug-and-Play priors for model based reconstruction, 2013 IEEE Global Conference on Signal and Information Processing, 2013,945-948. |
| [57] | J. Y. M. Y. Xiang Dong Zhang Li, Design of a magnetic induction tomography system by gradiometer coils for conductive fluid imaging, IEEE Access, 7, 56733-56744 (2019) · doi:10.1109/ACCESS.2019.2914377 |
| [58] | J. Zhang and B. Ghanem, ISTA-Net: Interpretable optimization-inspired deep network for image compressive sensing, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, (2018). |
| [59] | K. Zhang, W. Zuo, S. Gu and L. Zhang, Learning deep CNN denoiser prior for image restoration, Proc. IEEE Comput. Vis. Pattern Recognit., (2017), 3929-3938. |
| [60] | K. Zhang, W. Zuo and L. Zhang, Deep plug-and-play super-resolution for arbitrary blur kernels, Proc. IEEE Comput. Vis. Pattern Recognit., (2019), 1671-1681. |
| [61] | D. Zoran and Y. Weiss, From learning models of natural image patches to whole image restoration, in IEEE International Conference on Computer Vision, (2011), 479-486. |
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.
