Video Deblurring with GAN

This paper proposed A new video deblurring method based on generative adversarial network and Markov discriminant network which is to solve the problem of video blur caused by camera shaking or target movement. We combined the loss function based on pixel and feature space, and designed a new discriminant network based on Markov discriminant network, which facilitated the learning of image texture details by the network, resulting in a significant improvement in the quality of the generated images. We have made qualitative and quantitative comparisons between the method in this paper and similar methods on the test set and real video set. The experimental results show that the deblurred image by the method in the paper has a higher peak signal-to-noise ratio and richer detailed information.