Semantic-Aware Infrared and Visible Image Fusion

Aiming at the problems of poor visual effect and lack of background details in infrared and visible image fusion, we proposed an image fusion algorithm based on semantic segmentation. This method obtains the position and shape of each target in the source image through semantic segmentation, and we will set the weight value for each target, so that the information of the image can be preserved to a large extent. In addition, we also design a generative adversarial network, which uses different loss functions to adjust the generator and discriminator to ensure that the fused image is clearer and has richer texture features. Experimental results show that our method is superior to the new method in both visual effect and qualitative index.