Infrared and visible image fusion based on relative total variation decomposition

Abstract

In this study, we propose an infrared and visible image fusion method based on relative total variation decomposition, which can maintain the contrast information and texture information of source images simultaneously. Firstly, the source images are decomposed into structural layers and texture layers according to the relative total variation. The former is mainly the large frame structure and brightness of the source images, and the latter is the texture and noise with small gradient values. Secondly, the fusion weights are separately constructed according to characteristics of the structure and texture layers. The weights of structure layers are obtained according to the image energy to retain the brightness of source images, and the weights of texture layers are calculated according to the saliency map to retain the texture information. Finally, the fused image could be reconstructed according to previously obtained sub-images and weights. We also conduct qualitative and quantitative experiments on public datasets to verify the effectiveness of the proposed method. The results show that the proposed fusion method has leading advantages in maintaining contrast, avoiding edge blurring and reducing noise compared with several more advanced algorithms, and the fusion result is more in line with human vision.