Abstract
In recent years, infrared and visible image fusion task plays an important role in military and civilian applications. In this paper, a novel infrared and visible fusion method is proposed based on VGGNet and Visual Saliency Map to deal with the problem that loss of detail information and low brightness and contrast after traditional infrared and visible. Firstly, the source images are decomposed into low-frequency subbands and high-frequency subbands by Weighted Least Squares(WLS). Then, low-frequency subbands are fused by adopt an improved VSM weighted average strategy for adaptive fusion. While the high-frequency subbands are feed to the VGG-19 neural network to extract multi-layer features to generate weight map, and the high-frequency fusion are obtained by combining the l1-norm and the max selection strategy. Finally, the fused image will be reconstructed by combining the fused low-frequency subbands and high-frequency subbands parts. The experimental results show that the proposed fusion framework performances better than other typical fusion methods in both visual quality and objective evaluation.