基于Delaunay三角形网格的多视点视频编解码算法  

Multi View Video Coding and Decoding Algorithm Based on Delaunay Triangular Grid

作　　者：李娜 田文文 Li Na;Tian Wenwen(Department of Computer Science and Technology,Xi'an Jiaotong University City College,Xi'an 710018,China)

机构地区：[1]西安交通大学城市学院计算机系,西安710018

出　　处：《科技通报》2023年第11期19-23,共5页Bulletin of Science and Technology

基　　金：陕西省教育厅专项科研计划项目(22JK0444);西安交通大学城市学院2021年度校级科研项目(青年项目)(2021Q01)。

摘　　要：为解决多视点视频的三维视觉感知冗余及传输和存储数据量庞大等问题,本文提出基于Delaunay三角形网格的多视点视频编解码算法。将视频内的每帧图像作为原始参考图像,基于Delaunay将其自适应地分割成若干三角形,将分割后的三角形的各个顶点信息反馈至输出图像中,并基于Delaunay模型、参数仿射变换估计三角形中各点的视差矢量,得到多视点视频残差图像。运用EDCT(embedded discrete cosine transform)算法对残差图像进行编码,通过将熵编码的过程反向操作实现多视点视频解码。实验验证了所提算法可以完美地还原原视频内容,编解码后的视频情景逼真、画质更清晰,观感体验更佳。In order to solve the problem of 3D visual perception redundancy of multi view video and huge amount of data transmission and storage,a multi view video codec algorithm based on Delaunay triangle grid is proposed.Take each frame of the video as the original reference image,adaptively segment it into several triangles based on Delaunay,feed back the information of each vertex of the segmented triangle to the output image,and estimate the parallax vector of each point in the triangle based on Delaunay model and parameter affine transformation to obtain the multi view video residual image.EDCT(embedded discrete cosine transform)algorithm is used to encode the residual image,and multi view video decoding is realized by reversing the process of entropy encoding.The experiment verifies that the proposed algorithm can perfectly restore the original video content,and the video scene after encoding and decoding is lifelike,with higher picture quality and better visual experience.

关 键 词：DELAUNAY三角形网格 多视点视频 残差图像 矢量估计 编解码算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]