Infrared and visible image fusion based on non-subsampled contourlet transform

Abstract:
Image fusion is widely used in computer vision, robotics, military target recognition, and other fields. The fusion of infrared and visible images shows more information on a single image and highlights the target information, which is challenging due to the disadvantages of traditional fusion of infrared and visible images such as information loss, poor contrast, and non-translational invariance. In this paper, we propose an effective image fusion method to enrich the detail information of fused images. The visible and infrared images are decomposed using an improved non-subsampled contourlet transform method, and the high-frequency subbands are fused using large neighboring pixel differences, while the low-frequency subbands are fused using a combination of weighted averaging and absolute value extraction. Compared with previous solutions, our designed algorithm has translation invariance and the fusion rules retain the information of the low-frequency sub-band images, while better highlighting the texture details of the high-frequency sub-band images and improving the contrast. Extensive experimental results show that our improved fusion algorithm significantly exhibits better performance compared to the conventional algorithm.