V-SWIR-IF: Visible and Short-Wave Infrared Image Fusion

This paper presents an algorithm for the fusion of visible light and short-wave infrared (SWIR) images. The algorithm is structured around three core processes: image registration, fusion, and color reconstruction. To register multimodal images, an intensity-based image registration method is employed, which utilizes Mattes mutual information as metrics to guide the 1+1 evolutionary algorithm optimizer. This step produces a registered SWIR image. The image fusion step entails separating the visible image into three images by separating the RGB channels. The registered SWIR image is then fused with each of the three images individually using the Discrete Wavelet Transform (DWT). Finally, the RGB channels of the three fused images are reconstructed to complete the fused image’s color information. Experimental results on our dataset of visible and SWIR images show that the proposed algorithm achieves good performance in terms of image fusion, with a fast fusion speed of approximately 1 seconds per pair on a common household computer.