Patch Matching-Based Multitemporal Group Sparse Representation for the Missing Information Reconstruction of Remote-Sensing Images

Poor weather conditions and/or sensor failure always lead to inevitable information loss for remote-sensing images acquired by passive sensor platforms. This common issue makes the interpretation (e.g., target recognition, classification, change detection) of remote-sensing data more difficult. Toward this end, this paper proposes to reconstruct the missing information of optical remote-sensing data by patch matching-based multitemporal group sparse representation (PM-MTGSR). In the framework of sparse representation, the basic idea is to utilize the local correlations in the temporal domain and the nonlocal correlations in the spatial domain. Based on image patches, the local correlations are first taken into consideration. The similar patches are then grouped for joint sparse representation so that the nonlocal correlations are also considered. Owing to the patch matching of similar patches, the nonlocal correlations in the remote-sensing images are efficiently exploited. Simulated and real-data experiments demonstrate that the proposed method is effective both qualitatively and quantitatively.