Survey on Video Super-resolution Based on Deep Learning

Video super-resolution is the process of recovering its corresponding high-resolution video frames from a given low-resolution video sequence. In recent years, VSR has been driven by deep learning In order to further promote the development of VSR, the VSR algorithms based on deep learning are classified, analyzed and compared in this paper. First, according to the network structure, the existing methods are divided into two categories, namely the iterative network-based VSR and VSR based on recurrent network, and compare and analyze the advantages and disadvantages of different network models. Then, the VSR data set is comprehensively introduced, and the existing algorithms are summarized and compared on some commonly used public data sets. Finally, the The key issues in the VSR algorithm are analyzed, and its application prospects are prospected.