Infrared image enhancement based on an aligned high resolution visible image

Since the visual quality of an infrared (IR) image is usually unsatisfactory due to blurred edges and lack of textures, it is sometimes hard to obtain sufficient information from the IR image. In this paper, we present a novel framework for the IR image enhancement with the help of its aligned high resolution visible image. In the algorithm, we first prepare an aligned pair of IR and visible images through multi-sensor image registration. We then define a weight map based on edge correspondence in order to properly transfer the sharp edge property in the visible image to the IR image while avoiding unwanted blurring and distortion. We enhance the IR edges with high weights by applying visible-image-driven anisotropic diffusion with adaptive diffusion parameters. Finally, we deblur the remaining area to obtain a result that is enhanced uniformly over the whole IR image. Experimental results show that the proposed method can provide quality-improved IR images.