Infrared and visible image fusion based on BEMSD and improved fuzzy set

This study proposes a new fusion framework to solve the problem of weak self-adaptation in traditional multiscale transformation. A non-subsampled shearlet transform (NSST) decomposes the source image in multiscales and multidirections through a non-subsampled pyramid and shearlet transform. There is a problem in that the filter bank needs to be designed independently. The characteristics of the source images cannot be fully utilized. Bidimensional empirical mode decomposition (BEMD) can obtain a set of intrinsic mode functions (IMFs) and a residue according to the characteristics of the image itself. The high-frequency and low-frequency information decomposed by BEMD are more suitable to the source image features. Thus, using BEMD instead of the non-subsampled pyramid in an NSST allows the fusion framework to analyze the characteristics of the image with better multiscale and multidirection adaptability. For the proposed fusion framework, the corresponding fusion rules are needed. In this study, an improved fuzzy set is used as the low-frequency fusion rule, A contrast analysis combined with Euclidean distance is used as the high-frequency fusion rule. The proposed fusion rules are designed to enhance the adaptability while preserving the details of the source image as much as possible. The experimental results demonstrate that compared with the traditional method, the fusion results obtained by the proposed method are more realistic, and the infrared targets are more evident. Furthermore, the objective evaluation metrics are better.