A fast intra-frame coding method for 3d-hevc depth video

Abstract

The present invention discloses a fast encoding method within a 3D-HEVC depth video frame. Before encoding, the depth video frame is smoothed, that is, preprocessed, which reduces the complexity of the space domain and improves the encoding performance and virtual viewpoint quality; The preprocessed depth video frame is smoother, and the maximum division depth of more maximum coding units is concentrated in 0 and 1, so the BP neural network is used to actively learn and predict the maximum division depth of the maximum coding unit, and then through the corresponding color video frame The maximum division depth of the largest coding unit in the depth video frame and the edge texture information of the coding unit in the depth video frame are assigned to correct the predicted maximum division depth, and the division of the coding unit is terminated in advance. At the same time, according to the statistical results of the prediction mode of the coding unit of each layer and The predicted maximum division depth of the LCU determines the execution range of the prediction mode of each layer of the CU in advance, and finally improves the coding rate-distortion performance while greatly reducing the coding complexity.

Description

technical field [0001] The present invention relates to a coding technology of a multi-viewpoint video signal, in particular to a fast intra-frame coding method of a 3D-HEVC depth video. Background technique [0002] With the continuous advancement of video technology, 3D (Three-dimensional) video technology has been widely used in many aspects such as consumer entertainment, video conferencing and telemedicine. At present, a Free Viewpoint Video (FVV) system based on Multiview Video plus Depth (Multiview Video plus Depth, MVD) is the main implementation of 3D video. In the FVV system, the audience can freely choose the viewing angle to better obtain an immersive visual experience, and in order to achieve this goal, it is necessary to obtain enough viewpoints. In the MVD format video, the existing real viewpoints can be used to obtain enough virtual viewpoints through Depth Image Based Rendering (DIBR) technology, in which the depth video represents distance information, an...

AI Technical Summary

Problems solved by technology

Statistics show that the encoding time of 3D-HEVC depth video is about 10 times that of color video, which is unreasonable for depth video with simpler texture

[0004] At present, many scholars have proposed methods to reduce the complexity of 3D-HEVC depth video intra-frame coding, but these methods are focused on reducing the execution of DMMs or proposing new depth video intra-frame prediction modes to replace DMMs, these methods reduce 3D-HEVC The intra-frame encoding time of depth video is limited, mainly because these methods do not consider that depth video is relatively simple, and the optimal CU depth is mainly concentrated in 0 and 1.

For the estimated depth video, it is not processed in advance before encoding to improve encoding and rendering performance

Images