Fusion of infrared intensity and polarization images using embedded multi-scale transform

Multi-scale transforms are considered to be the effective methods in image fusion. However, infrared intensity and polarization images fusion based on these methods frequently arise the issues such as limited contrast ratio improvement and marginal area distortion. The embedded multi-scale decomposition has been proposed to integrate the advantages of different methods in this paper, where the low-frequency image continues to be decomposed by another multi-scale transform. The detailed procedure is shown as the following. Firstly, the low-frequency components and support value image sequences of infrared intensity and polarization images are obtained respectively with support value transform (SVT). Secondly, multi-scale bright and dim information are extracted from the last layer of the low-frequency images using the multi-scale top-hat decomposition, respectively. Thirdly, the resulted images are respectively fused and enhanced, and then fused with the two low-frequency images to get the fused low-frequency image. After that, the image is reversely transformed with the support value image sequences fused by extracting maximum gray to get the final image. Compared with the simple SVT and the multi-scale top-hat decomposition, the method suggested successfully improves the contrast ratio, the distortion of marginal area and the local coarseness. The validity of the method proposed is proven.