An efficient approach for spattotemporal image fusion with application to HSHT land cover change simulation

Monitoring fast land-cover changes at fine special resolution is an important requirement for global environmental change research. In this paper, we present a novel spatiotemporal fusion framework to efficiently simulate surface reflectance with both high-spatial and high-temporal (HSHT) resolution. The main contribution may be divided into two parts. First, a semi-physical unmixing algorithm is developed to super-resolve the heterogeneous landscapes with phenology change. Second, a super-resolution method is proposed for the land-cover-type changed landscape based on nonlocal-crossing-similarity property. The method is demonstrated on two types of data: images primarily with phenology changes and images primarily with land-cover-type changes. By comparing with other well-known spatiotemporal fusion algorithms, we evaluate the precision of the proposed approach.