Infrared and visible image fusion using structure-transferring fusion method

摘要

It is commonly believed that the purpose of the image fusion is to merge as much information, such as contour, texture and intensity distribution information from original images, as possible into the fusion image. Most of the existing methods treat different source images equally with certain feature extracting operation during the fusion process. However, as for the infrared (IR) and visible image fusion problem, the features of images taken from two imaging devices with different sensitive wave bands are different, sometimes even adverse. We can't extract and preserve the opposite information at the same time. To keep the targets salient in clutter background and visual friendly, in this paper, a novel IR and visible image fusion method called structure transferring fusion method (STF) is first proposed. Firstly, the structure-transferring model is built to transfer the grayscale structure from the visible input image into the IR image. Secondly, infrared detail enhancing strategy is carried out to supplement the missing details of the IR image. Experimental results reveal that the proposed STF method is both effective and efficient for IR and visible image fusion. The final fusion image with conspicuous targets and vivid texture is conducive to night vision surveillance for human observers.