Atomic Information Entropy Selection Based Generalized Orthogonal Matching Pursuit Algorithm

To solve the problem of atom selection errors in low signal-to-noise ratio environment and large-scale data processing by generalized orthogonal matching pursuit algorithm, a new atomic information entropy selection based gOMP algorithm is proposed. Firstly, the entropy of each atom, in the support set, is calculated one by one in the current iteration. Then, compared with the entropy value of current atom with the preset entropy value and determined whether to eliminate the atom. By the above iteration steps, the reconstruction accuracy of sparse signal is effectively enhanced due to the atoms of the support set are constantly being updated and become more reasonable. The experiment results prove that the proposed algorithm has better reconstruction ability for one-dimensional sparse signal or two-dimensional image when compared to the gOMP.