A regularized deep matrix factorized model of matrix completion for image restoration

It has been an important approach of using matrix completion to perform image restoration. Most previous works on matrix completion focus on the low-rank property by imposing explicit constraints on the recovered matrix, such as the constraint of the nuclear norm or limiting the dimension of the matrix factorization component. Recently, theoretical works suggest that deep linear neural network has an implicit bias towards low rank on matrix completion. However, low rank is not adequate to reflect the intrinsic characteristics of a natural image. Thus, algorithms with only the constraint of low rank are insufficient to perform image restoration well. In this work, we propose a Regularized Deep Matrix Factorized (RDMF) model for image restoration, which utilizes the implicit bias of the low rank of deep neural networks and the explicit bias of total variation. We demonstrate the effectiveness of our RDMF model with extensive experiments, in which our method surpasses the state of art models in common examples, especially for the restoration from very few observations. Our work sheds light on a more general framework for solving other inverse problems by combining the implicit bias of deep learning with explicit regularization.

[1]  Jicong Fan,et al.  Matrix completion by deep matrix factorization , 2018, Neural Networks.

[2]  Yin Zhang,et al.  Solving a low-rank factorization model for matrix completion by a nonlinear successive over-relaxation algorithm , 2012, Mathematical Programming Computation.

[3]  Zhao Kang,et al.  Top-N Recommender System via Matrix Completion , 2016, AAAI.

[4]  Yigang Cen,et al.  Rank adaptive atomic decomposition for low-rank matrix completion and its application on image recovery , 2014, Neurocomputing.

[5]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[6]  Yonina C. Eldar,et al.  Phase Retrieval via Matrix Completion , 2011, SIAM Rev..

[7]  Sanjeev Arora,et al.  Implicit Regularization in Deep Matrix Factorization , 2019, NeurIPS.

[8]  Mingming Yang,et al.  A novel patch-based nonlinear matrix completion algorithm for image analysis through convolutional neural network , 2020, Neurocomputing.

[9]  Nathan Srebro,et al.  Implicit Regularization in Matrix Factorization , 2017, 2018 Information Theory and Applications Workshop (ITA).

[10]  L. Rudin,et al.  Nonlinear total variation based noise removal algorithms , 1992 .

[11]  Martin Vetterli,et al.  The effective rank: A measure of effective dimensionality , 2007, 2007 15th European Signal Processing Conference.

[12]  P. Moradi,et al.  A novel collaborative filtering model based on combination of correlation method with matrix completion technique , 2012, The 16th CSI International Symposium on Artificial Intelligence and Signal Processing (AISP 2012).

[13]  Hao Wu,et al.  Image completion with multi-image based on entropy reduction , 2015, Neurocomputing.

[14]  Wei Li,et al.  Non‐Local Image Inpainting Using Low‐Rank Matrix Completion , 2014, Comput. Graph. Forum.

[15]  Yanfei Wang,et al.  A non-convex algorithm framework based on DC programming and DCA for matrix completion , 2014, Numerical Algorithms.

[16]  Emmanuel J. Candès,et al.  A Singular Value Thresholding Algorithm for Matrix Completion , 2008, SIAM J. Optim..