In this paper, we propose an infrared and visible image fusion network based on RepVGG architecture. This network adopts an encoder-decoder structure. The encoding network, which contains five RepVGG blocks, is utilized to extract deep features of infrared and visible images. Each layer of RepVGG blocks is constructed with 3x3, 1x1 and identity branches while training and converted to single-branch architecture constructed with 3x3 convolutional layers while inferring. These extracted features are added and the fusion image is reconstructed by the decoding network. The proposed method was compared with seven fusion methods and the result shows that the proposed fusion method can retain more contour and texture information with less noise. The proposed method is superior to the comparison methods. The code of the proposed fusion network is available at https://github.com/xiongzhangzzz/repvggfuse.
Infrared simulation software is widely used in the evaluation and model training of infrared equipment. Subject to foreign restrictions, domestic infrared simulation software hardly can meet the current requirements. However, visible simulation software is relatively mature. Thus, generated infrared images from visible images could be another way to replace infrared simulation software. In this paper, a deep learning style transfer method, cycle generative adversarial network, was introduced into the infrared simulation field. We captured visible and infrared images and extracted water surface targets to establish visible-infrared datasets. The cycle generative adversarial network model was applied to generate infrared simulation images from visible images. In terms of visual effect, generated infrared images were close to infrared images. To more holistically evaluate the visual quality of our results, we employ two tactics, structural similarity, and image quality. The result shows that the distribution of generated infrared images was well-matched to the origin visible images. And the image definition and details of generated infrared images were close to infrared images.