Infrared with visual image fusion based on two-scale decomposition and sturdy guided filtering technique

Abstract

Several Infrared (IR) and Visual (VIS) image fusion techniques have been widely used to acquire a novel image which may characterize the image accurately, completely and reliably. This process can serve an essential part in image processing applications. In this article, an enhanced IR and VIS image fusion technique is proposed by Two-Scale Decomposition (TSD) and Sturdy Guided Filtering (SGF) together to further increase the robustness of fusion process. Initially, IR and VIS images are decomposed for creating the base and detail layers. Then, Phase Congruency (PC) and Sum Modified Laplacian(SML) are applied to get saliency maps of base and detail layers, respectively. Also, Iteratively Reweighted Least Squares (IRLS) algorithm with GF, namely SGF is included instead of GF method in an efficient manner to properly smooth the weighting maps by preserving the depth edges that correspond to weak color edges and small structures. In this SGF technique, Enhanced Preconditioned Conjugate Gradient (EPCG) method is applied to optimize the RLS iteratively and select the conjugate paths for each iteration efficiently. This SGF can achieve high convergence rate and handle the structure inconsistency while properly preserving the edges. Experimental outcomes exhibit that the proposed TSD-PS-SGF based image fusion technique has higher performance over state-of-the-art techniques in terms of image feature-based, information theory-based and image structure-based metrics.