Sparse Representation-Based Image Fusion for Multi-Source NDVI Change Detection

Abstract

The normalized differential vegetation index (NDVI) is a useful index for change detection in remote sensing vegetation analysis. Multi-source NDVI change detection, which utilizes the NDVI information at different time from multiple satellites, can solve the problem of long-revisiting periods for a single source (satellite). However, the spatial resolution of NDVI calculated from the multispectral images of different satellites is different. A sparse representation-based image fusion method is proposed to improve the spatial resolution of NDVI. First, a high spatial-resolution vegetation index (HRVI) is utilized. The proposed method is based on the assumption that NDVI and HRVI with different resolutions will have the same sparse coefficients under some specific dictionaries. In the experiment, the proposed method is compared with several state-of-the-art methods to demonstrate its efficiency. Furthermore, its application in multi-source NDVI change detection verified by datasets from GF-1 and GF-2 satellites.