Image fusion with double sparse representation in wavelet domain

Abstract

Aiming at the problem of image fusion method based on sparse representation being easy to lose image details, a fusion method based on double sparse representation in wavelet domain is presented. Firstly, training images are transformed into the wavelet domain and learning dictionary for each sub-band respectively. And the double sparse representation coefficients for source images can be acquired by the learned dictionary and the coefficients being combined with the choose-max fusion rule. Finally, the fusion image is reconstructed by the inverse wavelet transform. The computer simulation results show that the proposed method performs very well in fusion both noiseless and noisy situations, and outperform conventional methods in terms of visual effect and quantitative fusion evaluation indexes.