Morphological EMD Based Image Fusion Using Image Statistics

Electronics and Communication Engineering

This work introduces a new and quick method for image fusion using morphological filters that is based on empirical mode decomposition (EMD). First, we create a multi-channel, bi-dimensional EMD technique for image fusion that is based on morphological filters. It can divide the input source images into many intrinsic mode functions (IMFs) with varying scales and a residue. It computes the upper and lower envelopes of a multi-channel image in the sifting processing using the morphological expansion and erosion filters. For multi-channel images, it considerably increases the calculation efficiency of EMD while preserving the decomposition quality. Second, rather than using the pixel-based fusion method typically employed in EMD-based image fusion, we employ a patch-based fusion strategy with overlapping partition to fuse the IMFs and residue. In this method, an image statistics-based weighted average rule is designed to fuse the IMFs, and the feature information extracted by IMFs is used as a guide to merge the residue. In addition to effectively extracting the significant information from the source images, this technique can lessen the spatial artifacts caused by the noisy pixel-wise maps. Numerous comparison tests on the fusing of multiple widely-used image data sets containing multi-focus and multi-modal images demonstrate that our recently suggested fusion method can get significantly better outcomes than the current EMD-based image fusion techniques. It also faces intense competition in terms of visualization, objective metrics, and time performance from the most advanced image fusion techniques.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.