Learned Convolutional Sparse Coding

TL;DR

This paper introduces a convolutional recurrent sparse auto-encoder that learns task-specific sparse convolutional dictionaries efficiently, achieving competitive image denoising and inpainting results with faster computation.

Contribution

It presents a novel convolutional recurrent sparse auto-encoder combining a sparse encoder based on LISTA and a convolutional decoder for efficient sparse coding.

Findings

Achieved competitive image denoising results to KSVD.

Performed well in image inpainting tasks.

Operates with significantly reduced run-time.

Abstract

We propose a convolutional recurrent sparse auto-encoder model. The model consists of a sparse encoder, which is a convolutional extension of the learned ISTA (LISTA) method, and a linear convolutional decoder. Our strategy offers a simple method for learning a task-driven sparse convolutional dictionary (CD), and producing an approximate convolutional sparse code (CSC) over the learned dictionary. We trained the model to minimize reconstruction loss via gradient decent with back-propagation and have achieved competitive results to KSVD image denoising and to leading CSC methods in image inpainting requiring only a small fraction of their run-time.