Fully Trainable and Interpretable Non-Local Sparse Models for Image Restoration
Preprint
Abstract
Non-local self-similarity and sparsity principles have proven to be powerful priors for natural image modeling. We propose a novel differentiable relaxation of joint sparsity that exploits both principles and leads to a general framework for image restoration which is (1) trainable end to end, (2) fully interpretable, and (3) much more compact than competing deep learning architectures. We apply this approach to denoising, jpeg deblocking, and demosaicking, and show that, with as few as 100K parameters, its performance on several standard benchmarks is on par or better than state-of-the-art methods that may have an order of magnitude or more parameters.Keywords
All Related Versions
This publication has 24 references indexed in Scilit:
- Deep unfolding of a proximal interior point method for image restorationInverse Problems, 2020
- Pointwise Shape-Adaptive DCT for High-Quality Denoising and Deblocking of Grayscale and Color ImagesIEEE Transactions on Image Processing, 2007
- Image Denoising Via Sparse and Redundant Representations Over Learned DictionariesIEEE Transactions on Image Processing, 2006
- Image denoising using scale mixtures of gaussians in the wavelet domainIEEE Transactions on Image Processing, 2003
- An EM algorithm for wavelet-based image restorationIEEE Transactions on Image Processing, 2003
- A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics2002
- Color plane interpolation using alternating projectionsIEEE Transactions on Image Processing, 2002
- A Wavelet Tour of Signal Processing1999
- Sparse coding with an overcomplete basis set: A strategy employed by V1?Vision Research, 1997
- Nonlinear total variation based noise removal algorithmsPhysica D: Nonlinear Phenomena, 1992