Intra-frame prediction method based on generative adversarial network

Abstract

The invention discloses an intra-frame prediction method based on a generative adversarial network, and the method comprises the steps: collecting an original video frame image, carrying out the resampling, and constructing an image data set; converting an image in the image data set into a YUV format, and performing size cutting processing on the image after format conversion to obtain a trainingdata set; training the generative adversarial network by taking the image of the training data set as the input of the generative adversarial network; embedding a generator part of the trained generative adversarial network into HEVC reference software; selecting a prediction reference block and inputting the prediction reference block into the trained generative adversarial network, outputting aprediction value, obtaining a prediction result by using an intra-frame prediction mode of an HEVC coding standard, comparing the rate distortion costs of the two modes, and selecting an optimal intra-frame prediction mode; and generating a prediction mode flag bit in the code stream according to the optimal intra-frame prediction mode, and continuously encoding according to the original HEVC standard encoding mode. According to the method, accurate prediction can be provided for complex textures and bent edges of the image.

Description

technical field [0001] The present invention relates to the field of video technology, and more specifically, to an intra prediction method based on a generative confrontation network. Background technique [0002] Video coding technology is committed to compressing video to obtain a code stream that is easy to transmit, especially for high-definition video transmission under limited network bandwidth. In video coding and decoding, prediction methods such as intra-frame and inter-frame can effectively reduce the spatial and temporal redundancy of video. The traditional video intra-frame prediction method is based on the preset that the texture in the image is directional, and uses several predefined fixed modes. When predicting, each intra-frame mode is enumerated one by one to select the best encoding cost. one of. This intra-frame prediction method can predict the texture structure of monotonic angle more effectively. However, based only on the assumption of directional...

AI Technical Summary

Problems solved by technology

[0006] In order to overcome the problem that the intra-frame prediction method in the prior art cannot effectively deal with complex

Images