Improved sparse representation fusion rules based infrared and visible image fusion algorithm

Classical sparse representation based fusion algorithms always suffer from the problem of insufficient evaluation of spatial detail information in image blocks. Combining the L1 norm with the Weighted Sum of Eight Neighborhood-based Modified Laplacian (WSEML) operators. An infrared and visible image fusion algorithm based on improved sparse representation fusion rules is proposed in this paper to improve the retention of texture information in the fused image and the intelligibility of the fused image by providing a comprehensive measure of the spatial and transform domains of the image block’s activity level. The improved algorithm is then combined with Non-Subsampled Shearlet Transform (NSST), which is applied to the low-frequency part of the image after multi-scale decomposition, while the high-frequency part of the image is fused using the classical “maximum absolute” rule. At last, the inversed NSST is used to complete the image fusion. Experimental results show that the proposed algorithm: has advantages over traditional NSST, traditional sparse representation algorithms in effectively measuring and preserving low-frequency image details; has the advantages over traditional NSST algorithm, traditional sparse representation-based algorithms and other state-of-art algorithms in improvement in global fusion performance.