Cloud Removal in Hyperspectral Satellite Images Using Low-rank Tensor Completion

We propose an unfolding-based low-rank tensor completion (LRTC) algorithm for cloud removal in hyperspectral satellite images. We first formulate cloud removal as an LRTC-based joint optimization problem, incorporating handcrafted priors for hyperspectral image acquisition and implicit regularization functions to compensate for modeling inaccuracies. We then solve the optimization problem iteratively and develop a multistage deep unfolded network. In this network, each stage corresponds to an iteration of the iterative algorithm in which the optimization variables and regularizers are updated using closed-form solutions and learned deep networks, respectively. Experimental results demonstrate that the proposed algorithm achieves better restoration performance than state-of-the-art algorithms in both quantitative and qualitative comparisons.