VDFEFuse: A novel fusion approach to infrared and visible images

Abstract

To address the problems of energy loss and edge blur when fusing infrared and visible images with traditional methods, a novel infrared and visible image fusion approach based on the visual differentiation feature extraction is proposed, termed as VDFEFuse. Owing to the unique thermal radiation imaging characteristic of the infrared sensor, more intensity information is preserved in the infrared images. In this approach, we construct a visual differentiation feature extraction operator to compensate for the loss of such significant features in the fusion process to the greatest extent. Then, through the LatLRR-based multilevel decomposition and the specific fusion strategies, a set of fused sub-images is generated. To select the best fusion result from fused sub-images, a novel hierarchical structure decision model is formulated in which the qualitative and quantitative evaluations of all fused sub-images can be taken into account simultaneously. Compared with nine image fusion methods, experiments showed that not only does the proposed algorithm have better visual effects from subjective judgment and objective evaluation, but it also can better retain the rich brightness information in the source image.