Blind Hyperspectral-Multispectral Image Fusion via Graph Laplacian Regularization

TL;DR

This paper introduces a novel hyperspectral-multispectral image fusion method that leverages graph Laplacian regularization to achieve super-resolution without prior knowledge of spatial degradation, effectively handling misaligned data.

Contribution

The proposed algorithm uniquely estimates the blur kernel and fuses images iteratively without requiring prior degradation information or perfect alignment.

Findings

Improves fusion quality over existing methods.

Handles unknown spatial degradation effectively.

Demonstrates robustness on various datasets.

Abstract

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.