Pixel level jointed sparse representation with RPCA image fusion algorithm

This paper presents a pixel level image fusion technique based on integrating the sparse representation with robust principle component analysis algorithm (RPCA) to promote relevant information, eliminate noise and preserve edges. In this method, the input images are first decomposed into their common components and sparse or uncommon components using RPCA algorithm. Secondly, the sparse representation using OMP algorithm with a trained dictionary is adapted to fuse those two components by adding the common low rank sparse coefficient to the maximum norm of uncommon sparse coefficients. Finally, we reconstruct the fused image from the fused sparse coefficients and adaptive dictionary. This proposed image fusion technique is evaluated and compared with OMP based on visual inspection and quality measures. The conjoint OMP-RPCA simultaneous method shows promising results and best performance in a variety of different image types including medical, multifocus and infrared and visible light images.