3D-HEVC depth map coding unit division rapid decision-making method based on machine learning

Abstract

The invention discloses a 3D-HEVC depth map coding unit rapid decision-making method based on machine learning, which is used for solving the problem of high coding complexity caused by adding a depthmap and a new coding tool in 3D-HEVC through constructing a depth map fast coding unit (CU) level decision-making system based on depth gradient boosting (XGBoost). The 3D-HEVC depth map coding unitrapid decision-making method comprises two parts: XGBoost model training and fast CU segmentation decision. Data mining and machine learning are used, and a decision model is constructed by using texture information of a depth map as a feature attribute vector and whether a current CU continues to serve as a sub-class label to be divided into sub-CUs or not. Feature attributes are extracted from the encoding process, and a well trained model is used to determine whether the CU continues partitioning. Compared with a standard encoder, the 3D-HEVC depth map coding unit rapid decision-making method has a better performance improvement effect; and compared with related work, the encoding performance is improved to different degrees.

Description

technical field [0001] The present invention relates to a video coding technology based on 3D-HEVC, in particular to a fast decision-making method for dividing depth map coding units based on machine learning in 3D-HEVC coding Background technique [0002] With the rise of computer multimedia technology and the continuous development and improvement of video technology in film and television, digital video technology has become a research hotspot in modern academic and industrial circles. Three-dimensional video not only brings a higher sense of experience to people's senses, but also has a wide range of application prospects, such as medicine, education and other fields. In order to meet users' needs for viewing comfort and viewing freedom, 3D video application equipment is also gradually developing in the direction of providing more viewpoints, such as Autostereoscopic Display and Free Viewpoint Television, enabling users You can choose a certain point of view to watch. ...

AI Technical Summary

Problems solved by technology

[0010] The object of the present invention is to propose a machine learning-based fast decision-making method for depth map coding unit division in view of the high computational complexity of the depth map coding unit quadtree division method in the HEVC-based 3D video coding standard 3D-HEVC , under the premise of ensuring the quality of the video into a virtual viewpoint, it can effectively reduce the coding complexity, improve the coding efficiency, and shorten the coding time

Images