Image Fusion Based on Morphological Component Analysis via Gradient

Abstract

I306 fusion is a process to strengthen the human perception of individual source images from the same scene. We propose a novel method to address fusion based on morphological component analysis (MCA) and gradient comparison. We first use MCA to separate source images into multi-components (cartoon, texture and few residual images). Then computing the sum of gradients for every pixel in each image component and select suitable pixels in the fused image. In the experiments, the proposed method brings a state-of-the-art fused performance compared with traditional algorithms in terms of the visual perceptions and objective evaluations.