Video super-resolution recovery method and system based on bidirectional mask video super-resolution network

The invention relates to the technical field of video super-resolution and artificial intelligence algorithms, and discloses a video super-resolution recovery method and system based on a bidirectional mask video super-resolution network, and the method comprises the following specific steps: S1, constructing the bidirectional mask video super-resolution network of a random mask; degenerating the training video data set to obtain a low-resolution video data set; s2, inputting the low-resolution video data into a bidirectional mask video super-resolution network of a random mask, and carrying out Switch White based on a dark channel; s3, performing bidirectional video masking on the processed video data; s4, recovering the video features into a high-resolution video to obtain a high-resolution video data set; s5, calculating loss, and training a bidirectional mask video super-resolution network; and S6, carrying out super-resolution recovery on the real video. The method solves the problems of low training spe