Region-based image fusion method using bidimensional empirical mode decomposition

摘要

We propose a novel multiscale decomposition (MSD) im-nage fusion algorithm, which is region-based image fusion using bi-ndimensional empirical mode decomposition (BEMD). BEMD is a 2-Dndata-driven decomposition derived from the empirical mode decom-nposition (EMD), which does not require predetermined filter or wave-nlet function. The input images are decomposed into a number ofnintrinsic mode functions (IMFs) as well as a residual image. Thenfusion is performed region by region based on the segmentations ofnthe input images to produce composite BEMD representation, andnthen the inverse BEMD transform is applied to obtain the fused im-nage. Experiments show that the proposed image fusion algorithmnprovides superior performance over traditional fusion schemes innterms of both objective metrics and visual quality.