Fusion of hyperspectral and panchromatic images using structure tensor and matting model

Abstract The purpose of hyperspectral image fusion is to integrate the spectral information of hyperspectral (HS) image and the complementary spatial information of panchromatic (PAN) image. The result can improve the accuracy of subsequent image processing such as classification and detection. Component-substitution (CS)-based methods are popular HS image fusion approaches, which are efficient, and simple to implement. A novel CS-based hyperspectral image fusion framework is presented by combining structure tensor and matting model in this paper. The proposed scheme not only accurately estimates the missing spatial components of the HS images by using structure tensor, but also effectively preserves their spectral components by combining the advantages of structure tensor and matting model. Specifically, a structure-tensor-based adaptive weighted fusion strategy is proposed for generating the alpha channel. The high-resolution HS image is obtained by substituting the original alpha channel of the interpolated HS image with the generated alpha channel. Experimental results with synthetic datasets and real datasets prove the potential of the proposed algorithm in preserving spectral information and enhancing spatial information.