Multi-scale Bézier Filter Based Infrared and Visual Image Fusion

In this study, we have proposed a multi-scale Bézier filter based method for infrared and visual image fusion. Specifically, we first exploit our developed multi-scale Bézier filter to decompose the infrared image and visual image into a set of multi-scale bright feature maps, multi-scale dark feature maps and a base image, respectively. Then, each scale of bright feature maps and dark feature maps are respectively fused by selecting their elementwise-maximums, then amplified by a scale-related coefficient, and finally integrated as a fused bright feature map and a fused dark feature map. Afterwards, the two base images are fused by averaging their mean values and selecting the elementwise-maximums of their large-scale features. Finally, the fusion image is produced by integrating the fused bright feature map, dark feature map and base image. Extensive experiments demonstrate that our method significantly outperforms the state-of-the-art image fusion methods from both qualitative and quantitative aspects.