Double-Level Binary Tree Bayesian compressed sensing for block sparse image

Abstract

Based on the fact that some image signals possess the block sparsity in practical application environment, a novel Compressed Sensing (CS) algorithm for block sparse image is proposed in this paper. Namely, a Double-level Binary Tree (DBT) Bayesian model is proposed for the block sparse image at the same time the relationship of the root node and the leaf node of this DBT structure is defined as “genetic characteristic”. Then, the block clustering for the block sparse image can be executed successfully and effectively by utilizing Markov Chain Monte Carlo (MCMC) method. The simulation results prove that, our proposed method for the block sparse image signal can get better recovery results with less computation time.