Detail preserved fusion of visible and infrared images using regional saliency extraction and multi-scale image decomposition

Fusion method of infrared and visible images to a synthetic image is a significant research topic in image process. It is an effective way to extract the detail information of the visible image and target regions of the infrared image in the final fused result. In this paper, a detail preserved fusion algorithm of visible and infrared images is proposed based on the regional saliency extraction technique and multi scale image decomposition. The multi-scale image decomposition is firstly applied to the infrared and visible images under the image smoothing framework using L1 fidelity with L0 gradient. Then for each decomposition layer the saliency map is extracted by the frequency-tuned saliency map extraction algorithm. The final fused result is reconstructed by synthesizing different levels with proper weight values. Experiments are implemented to test the performance of the proposed fusion approach compared with other excellent fusion methods. Both the subjective perception and quantitative index results demonstrate that the proposed approach obtains better performance in preserving the edge detail information as well as enhancing the infrared targert signals.