Thick Cloud Removal in Multi-Temporal Remote Sensing Images via Frequency Spectrum-Modulated Tensor Completion

Clouds often contaminate remote sensing images, which leads to missing land feature information and subsequent application degradation. Low-rank tensor completion has shown great potential in the reconstruction of multi-temporal images. However, existing methods ignore different low-rank properties spatial temporal dimensions, such that they cannot utilize adequately. In this paper, we propose a new frequency spectrum-modulated method (FMTC). First, images are rearranged as third-order spatial–temporal tensors for each band. Then, Fourier transform (FT) is introduced dimension generate spatial–frequential tensor. view fact features represent low-frequency components fickle clouds high-frequency time domain, chose adaptive weights matrixes, according spectrum. Invert Transform (IFT) was implemented. Through method, joint constraint achieved. The simulated data experiments demonstrate FMTC applicable on land-cover types sizes. With real experiments, have validated effectiveness stability time-series image reconstruction. Compared with other algorithms, performance better quantitative qualitative terms, especially when considering spectral accuracy continuity.