Multiscale Feature Interactive Network for Multifocus Image Fusion

In deep learning (DL)-based multi-focus image fusion, effective multi-scale feature learning is a key issue to promote fusion performance. In this paper, we propose a novel DL model named Multi-Scale Feature Interactive Network (MSFIN), which can segment the source images into focused and defocused regions accurately by sufficient interaction of multi-scale features from layers of different depths in the network, for multi-focus image fusion. Specifically, based on the popular encoder-decoder framework, two functional modules, namely, multi-scale feature fusion (MSFF) and coordinate attention up-sample (CAU) are designed for interactive multi-scale feature learning. Moreover, the weighted binary cross entropy (WBCE) loss and the multi-level supervision (MLS) strategy are introduced to train the network more effectively. Qualitative and quantitative comparisons with 19 representative multi-focus image fusion methods demonstrate that the proposed method can achieve the state-of-the-art performance. The code of our method is available at https://github.com/yuliu316316/MSFIN-Fusion.