Blind Hyperspectral-Multispectral Image Fusion via Graph Laplacian Regularization
This addresses the challenge of enhancing image resolution in remote sensing applications, though it is incremental as it builds on existing fusion methods by handling unknown degradations.
The paper tackles the problem of fusing low-resolution hyperspectral and high-resolution multispectral images without prior knowledge of spatial degradation or perfect alignment, resulting in a super-resolution image with improved quality and accurate blur kernel estimation.
Fusing a low-resolution hyperspectral image (HSI) and a high-resolution multispectral image (MSI) of the same scene leads to a super-resolution image (SRI), which is information rich spatially and spectrally. In this paper, we super-resolve the HSI using the graph Laplacian defined on the MSI. Unlike many existing works, we don't assume prior knowledge about the spatial degradation from SRI to HSI, nor a perfectly aligned HSI and MSI pair. Our algorithm progressively alternates between finding the blur kernel and fusing HSI with MSI, generating accurate estimations of the blur kernel and the SRI at convergence. Experiments on various datasets demonstrate the advantages of the proposed algorithm in the quality of fusion and its capability in dealing with unknown spatial degradation.