Research On 3D Video Coding Optimization And Method

Nowadays,people are constantly pursuing higher audio-visual quality,which makes three-dimensional video that provide deep information-aware has received extensive attention.Multi-view plus depth 3D video,as the main data format of 3D video,has become a focus of attention and research in the industry and academia.There are more than one view for multi-view plus depth stereoscopic video,so its data is usually several times that of conventional two-dimensional video.In addition to the rapid development in the perspective range,spectral range,resolution,and dynamic range,the data is growing explosive,which makes a more effective 3D video compression technology an urgent need.The quality evaluation that can accurately measure the compression performance of an algorithm is an important prerequisite for realizing this demand.This article has researched part of these two issues.The main innovations and contributions are shown as follow.1.In the actual operation of quality evaluation,both the coding bit and the video quality of the video are changing.BD-Rate is widely used to measure the coding efficiency of a algorithm.This article finds and proves that in the process of calculating BD-Rate,when the PSNR is greatly different,it will introduce statistical errors to BD-Rate and cause inaccurate.On the other hand,due to the inter-view reference,the PSNR values of each texture view differ greatly in three-dimensional video coding.Therefore,this article shows that BD-Rate is prone to misalignment when measuring the video quality of total texture.To solve this problem,this article proposes an amendment of BD-Rate's statistics,which makes the BD-Rate indicator more accurate.2.In order to improve the coding efficiency of 3D video,this article models and quantifies the GOP-level reference relationship between the same view and various view in 3D video and uses their reference correlation to set the Quantization parameter offsets in 3D video to reduce redundant information.Experimental results show that the proposed method can improve the efficiency of 3D video coding.