A Robust Hyperspectral Image Fusion Method for Unregistered Data

Abstract

Fusion of hyperspectral image (HSI) and panchromatic image (PI) is an elementary yet crucial topics in remote sensing image processing. The fused data will be characterized by the high spatial and spectral information simultaneously. Although conventional pixel level based fusion methods have been proposed and accomplish the task, they ignore the condition for fusion that the input source images should be registered perfectly. To address the problem, we propose a novel fusion scheme, which aims at fusing HSI and PI for registered and unregistered cases. The proposed scheme first extracts spatial information from PI, and then embeds the spatial information into the HSI. In this way, the proposed method achieves competitive spatial detail and spectral information, which is also demonstrated in experiment results.