Robust Multifocus Image Fusion via Multiple Visual Features and Guided Filtering

This paper aims at modeling an effective activity measurement for focus areas detection in multifocus images. Considering that the existing methods lack efficiency and are not able to deal with mis-registered source images well, we propose a multiple visual features and guided filtering based fast and robust multifocus image fusion method. First, the initial fusion decision map is obtained by detecting the focus area in source images through measuring three visual features, which are contrast saliency, sharpness and structure saliency; then, the final fusion weight map is acquired by optimizing the initial decision map through morphological filtering and guided filtering to make full use of spatial consistency and to resist artifacts; finally, the fused image is obtained by weighted averaging the source images according to optimized weight map. Experimental results show that our method outperforms existing state-of-the-art multifocus image fusion algorithms, in terms of both subjective and objective quality assessments.