Rethinking the CSC Model for Natural Images

Abstract

Sparse representation with respect to an overcomplete dictionary is often used when regularizing inverse problems in signal and image processing. In recent years, the Convolutional Sparse Coding (CSC) model, in which the dictionary consists of shift-invariant filters, has gained renewed interest. While this model has been successfully used in some image processing problems, it still falls behind traditional patch-based methods on simple tasks such as denoising. In this work we provide new insights regarding the CSC model and its capability to represent natural images, and suggest a Bayesian connection between this model and its patch-based ancestor. Armed with these observations, we suggest a novel feed-forward network that follows an MMSE approximation process to the CSC model, using strided convolutions. The performance of this supervised architecture is shown to be on par with state of the art methods while using much fewer parameters.

Cited by in corpus (12)

• Deep Residual Autoencoders for Expectation Maximization-inspired Dictionary Learning
• CDLNet: Noise-Adaptive Convolutional Dictionary Learning Network for Blind Denoising and Demosaicing
• CDLNet: Robust and Interpretable Denoising Through Deep Convolutional Dictionary Learning
• Fully Trainable and Interpretable Non-Local Sparse Models for Image Restoration
• On the convergence of group-sparse autoencoders
• Deep learning based dictionary learning and tomographic image reconstruction
• Efficient Consensus Model based on Proximal Gradient Method applied to Convolutional Sparse Problems
• Discriminative reconstruction via simultaneous dense and sparse coding
• Learned Interpretable Residual Extragradient ISTA for Sparse Coding
• Improved Image Generation via Sparse Modeling
• Tensor Convolutional Sparse Coding with Low-Rank activations, an application to EEG analysis
• Convolutional dictionary learning based auto-encoders for natural exponential-family distributions