Multi-focus image fusion based on depth estimation in HSV space

A robust multi-focus image fusion method is proposed to generate an all-in-focus image with all objects in focus by merging multiple images. The proposed method first estimates local focus maps using a novel measure of Gaussian model combined with joint bilateral filtering in HSV space. Then, a propagation process is conducted to obtain accurate focus maps based on a traditional natural image matting model that makes full use of the spatial information. The fused all-infocus image is finally generated by a focus-selected strategy. Experimental results demonstrate that the proposed method has state-of-the-art performance for multi-focus image fusion under various situations encountered in practice, even in cases with little edge information.