Efficient fusion for infrared and visible images based on compressive sensing principle

摘要

In this study, the potential application of compressive sensing (CS) principle in the image fusion for infrared (IR) and visible images is studied. First, the theory of CS is introduced briefly. Some comparative analyses of different reconstruction techniques are carried out in view of their performance in multisensor image recovery and the minimum number of sampling measurements one has to take to achieve perfectly reconstruction of images is investigated afterwards. Then, a novel selfadaptive weighted average fusion scheme based on standard deviation of measurements to merge IR and visible images is developed in the special domain of CS using the better recovery tool of total variation optimisation. Both the subjective visual effect and objective evaluation indicate that the presented method enhances the definition of fused results greatly, and it achieves a high level of fusion quality in human perception of global information. On the other hand, no structure priori information about the original images is required and only some concise fusion computation of compressive measurements is needed in the authors' proposed algorithm, thus it has superiority in saving computation resources and enhancing the fusion efficiency.