Image Fusion based on Variational Method for Maintenance of Gradient and Intensity

Infrared and visible image fusion technology helps to improve the spatial resolution. It mainly preserves the features and details of the source images and generates a fusion image with visual enhancement. In this paper, based on the gradient features and intensity information of the source images, an optimization model for image fusion is built. Firstly, the pre-fused gradient of the source images is obtained by combining the structure tensor and the proposed local gradient similarity, where local gradient similarity is used to make the fused gradient direction more accurately. Secondly, the source images are reconstructed into salient and non-salient images according to the comparison of the pixel intensity. A weight map before the non-salient image in the optimization model makes the effective details preserved, so that the pre-fused images consist of the salient image and the non-salient image with a weight map. Finally, an optimization model is constructed to constrain the gradient and intensity of the final fused image close to the pre-fused gradient and the pre-fused images. The final fused image is obtained from solving the optimization model by use of the variational method. The experimental results are evaluated from subjective and objective assessments, which show the effectiveness of the proposed algorithm.