MCDFD: Multifocus Image Fusion Based on Multiscale Cross-Difference and Focus Detection

As an important multisensor data fusion technology, multifocus image fusion (MFIF) integrates all focused information from source images to provide a comprehensive and objective interpretation. However, current MFIF algorithms have limitations on blurred edges and oversharpening of the fused images, resulting in loss of details or introduction of artifacts. In this article, we propose a novel MFIF scheme based on multiscale cross-difference and focus detection. First, to extract the pixel information at different scales, we express the source images at different scales using multiscale-guided filtering. Subsequently, we proposed a multiscale cross-difference strategy for obtaining the salient map, which accurately highlights the pixels within the focused regions. Meanwhile, we designed an effective two-scale decision map generation model to accurately refine the focus decision maps. Finally, we introduce guided filter (GF) and consistency verification operations to optimize and obtain the final decision maps and the fused result. The proposed model can effectively address edge blur and detail loss and improve the clarity and visual quality of fused images. Experiments were performed on two popular public datasets and the subjective and objective evaluations indicate that the proposed method has a better performance than 28 state-of-the-art (SOTA) methods. The code of this work was available at
https://github.com/ixilai/MCFDF
.