The flexible signature dictionary

Abstract

Dictionary learning and Sparse representation of signals and images has been a hot topic for the past decade and aims to help find the sparsest representation for the signal(s) at hand. Typically, the Dictionary learning process involves finding a large number of free variables. Also, the resulting dictionary in general does not have a specific structure. In this paper we use the ideas from Image Signature Dictionary and General overlapping frames and proposed a flexible signature dictionary. We show that the resulting signatures capture the essence of the signal and can represent signals of their own type very well in opposed to signals of other types.