Method, system and terminal for video restoration by using deep convolutional neural network

Abstract

The invention provides a method, system and terminal for video restoration by using a deep convolutional neural network. The method comprises the steps of carrying out the preprocessing of an originalvideo data set, and forming a training set, constructing a feature extraction network model, constructing a loss function, training the constructed feature extraction network model by jointly utilizing the training set and the loss function, and adjusting parameters of the model according to a result generated by training to obtain a final video restoration model, and repairing the video by usingthe obtained video restoration model. The invention provides a method, a system and a terminal for video restoration by using a deep convolutional neural network. The defect that a traditional methoddepends on manual definition and feature extraction is overcome, the powerful feature extraction capacity of the deep convolutional neural network is utilized, edge information between frames servesas restoration guidance, the features of video frames in the space domain and the time domain are utilized as much as possible, and meanwhile the subjective and objective quality evaluation indexes ofvideo restoration are improved.

Description

technical field [0001] The present invention relates to the technical field of video restoration, in particular to a method, system, and terminal for video restoration using a deep convolutional neural network. Background technique [0002] In recent years, with the popularization of the Internet, the video business has developed prosperously, and people's pursuit of video quality is also getting higher and higher. However, at present, most of the video streams received by the client end are transmitted to the client end through channels after High Efficiency Video Coding (HEVC). As a result, the video code stream received by some users is lost, resulting in damage to the decoded video. In addition, it is often necessary to repair some precious videos that are damaged or whose channels are susceptible to interference, such as surveillance videos. Therefore, in the face of such visual picture damage, how to restore it through technical means is particularly important, and i...

AI Technical Summary

Problems solved by technology

[0007] 1. Excessive dependence on light field information, not sensitive to some grayscale information;

[0008] 2. The network structure is quite complex, and too much optical flow information needs to be tracked, which is not easy to train

[0010] 1. In the time domain, the advantages of the similarity of information between the front and back frames cannot be fully utilized;

[0011] 2. In the airspace, only a single image is considered for restoration, and the amount of information is too small

[0012] At present, there is no description or report of the similar technology of the present invention, and no similar data at home and abroad have been collected yet.

Images