Region Feature Based Multi-scale Fusion Method for Thermal Infrared and Visible Images

To overcome the defects of existing algorithms that the target information and edge details are easily lost and that fusion image contrast is low, a novel fusion method that combines region feature and multi-scale transform for thermal infrared and visible images is proposed in this paper. Firstly, the source infrared and visible images are segmented based on adaptive pulse coupled neural network (PCNN) and two-dimension Renyi entropy, and a joint segmentation map can be acquired through region joint operation. Then the original images are multi-scale and multi-directional decomposed by nonsubsampled contourlet transform (NSCT). After that, the fusion rules are designed based on region feature difference in NSCT domain. Finally, the fusion image is reconstructed by NSCT inverse transform. Experimental results show the proposed method can effectively fuse infrared target feature, preserve the background information as much as possible, and obtain good contrast. The proposed method is superior to the traditional methods in terms of both subjective evaluation and objective evaluation.