Multi-exposure Image Fusion using Patchbased Component Decomposition

Abstract

Multi exposure image fusion is always a challenge in task in image processing. The multiple images with the different image content, when mixed using a fusion formula generate different effects. One of the most prominent effects is ghosting effect. Ghost in effect occur even in capturing of images. The smallest ghosting effect may be treated as image blur. To handle ghosting effect as well as many other affects that are generated in the process of fusion are treated in the proposed technique. The proposal scheme introduces a completely new representation that may be explorer for the for many different applications. First the input images are decomposed into several patches. As the fusion involves multiple input images the special correlated patches are further grouped into a class. Individual patches of the class are decomposed into three logical components named strength structure and intensity. These components are calculated for all the patches of the class. Now using the rule of fusion these logical components are derived for the whole class. The decomposition of a patch into logical components is unique as well as invertible hence using the generated components patches restored. Simulation results prove the superiority of the scheme proposed.