Algorithm for image fusion based on orthogonal grouplet transform and pulse-coupled neural network

Abstract

Orthogonal grouplet transformation is a kind of weighted multiscale Haar transform. It can effectively approximate the geometric structure of any shape in a small region or one with long association. Pulse-coupled neural network (PCNN), a simplified neural network, is able to extract useful information from a complex background without learning or training. In order to get better fusion effect, grouplet transform is applied for its advantage of describing complex edges or texture, and then coefficients are fused by PCNN. An algorithm named GT-PCNN based on grouplet transform and PCNN is proposed. It defines a coefficient fusion rule based on fire times and an association field fusion rule based on regional fire variance. The image fused by GT-PCNN has clear edges and texture and the overall effect looks good. Indicators of average gray, entropy, and mutual information are all better than those of the average algorithm, the principal component analysis algorithm and the algorithm based on wavelet transform and PCNN.