基于光场渲染的多视点视频编解码方法研究  

The research of a multi-view video encoding and decoding method based on light field modeling

作　　者：程龙[1] 郭立[1] 陈晓琳[1] 袁红星[1] 

机构地区：[1]中国科学技术大学电子科学与技术系,安徽合肥230027

出　　处：《中国科学技术大学学报》2010年第8期812-817,共6页JUSTC

基　　金：教育部-英特尔信息技术专项科研基金(MOE-INTEL-08-08)资助

摘　　要：首先提出一种面向移动设备的多视视频纹理编码方法,编码端利用重建的3D网格信息对4D光场重采样,得到顶点的光场矩阵;然后提出一种基于奇异值分解(SVD)的自适应分解级数选取方法,保证真实感的同时尽可能减少光场数据量.为了配合解码端的图形硬件加速,采用了S3TC纹理压缩标准对分解后的矩阵进一步压缩,减少编码端数据量以及解码端纹理渲染的带宽需求.实验结果表明,在保证较好的渲染效果下,提高了压缩比,并且满足实时渲染要求.A method for video texture encoding and decoding on a mobile device was proposed. On the encoder side, the 4D light field was resampled based on the reconstructed 3D model and then the vertex light field matrix was acquired. An optimization was to adjust the decomposition level self-adaptively based on SVD to decrease the data quantity while guaranteeing the 3 dimentionality in light field matrix factorization. In order to utilize graphics hardware acceleration, S3TC texture compression standard was used to further decrease data quantity at the encoder end and bandwidth demand at the decoder end. The experiment results show that the proposed method can improve the compression rate while guaranteeing better rendering effect, and meet the demands of real-time rendering.

关 键 词：多视点视频 光场渲染 S3TC 自适应奇异值分解 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]