Hyperspectral and multispectral image fusion using dual-source localized dictionary pair

Abstract

Hyperpsectral image (HSI) provides abundant and detailed spectral information with limited spatial resolution. When a multispectral image (MSI) with higher spatial resolution of the same observed scene is available, this limitation on spatial resolution can be handled by applying fusion techniques. In this work, a novel HSI and MSI fusion approach using dictionary-based reconstruction is proposed. To incorporate more effective information for fusion, dual-source dictionary pair including a dictionary on low spatial resolution and a dictionary on high spatial resolution, is constructed using both HSI and MSI. Furthermore, to reduce the calculation cost, a localized strategy is applied instead of the global one. Finally, the fused result is reconstructed with dictionary using collaborative representation. Simulative experiments illustrate its outperformance over some state-of-the-art HSI and MSI fusion approaches.