A novel Matching Pursuit algorithm with adaptive subdictionary

Abstract

Sparse decomposition of a signal can be obtained by decomposing the signal in an over complete dictionary. Matching Pursuit (MP) is a greedy algorithm to realize such decompositions, and has been widely studied in many signal processing researches. Through exploiting the structure shared by these so-called adaptive decomposition dictionaries, we propose a novel MP algorithm which can reduce computation and storage requirement dramatically.