Taj-Shanvi Framework for Image Fusion Using Guided Filters

Abstract

Multi-focus image fusion aims to produce an all-in-focus image by integrating a series of partially focused images of the same scene. A small defocused (focused) region is usually encompassed by a large focused (defocused) region in the partially focused image; however, many state-of-the-art fusion methods cannot correctly distinguish this small region. To solve this problem, a novel Taj-Shanvi framework, used for multi-focus image fusion algorithm based on multi-scale focus measures and generalized random walk (GRW), is implemented. First, multi-scale decision maps are obtained with multi-scale focus measures. Then, multi-scale guided filters are used to make the decision maps accurately align with the boundaries between focused and defocused regions. Next, GRW is used to combine these decision maps at different scales. After obtaining them, these maps are aligned using the watershed technique, whose edges are further smoothed using the guided filter. Experimental results are obtained by using few quality parameters, namely, entropy, edge structure-based similarity index measure, spatial frequency, mutual information, and so on, to evaluate the quality of the final fused image. Quality parameter assessment demonstrates that the proposed method produces a better quality fused image than conventional image fusion techniques.