HEVC兼容的全景视频运动补偿预测算法

发布时间:2017-12-11 作者:胡强,马文广,石志儒 阅读量:

[摘要] 利用图像渲染技术,全景视频能够为终端用户提供沉浸的虚拟现实体验。不同于传统二维视频,360°全景视频不仅包含平移运动还包括仿射运动,后者主要由于球面映射所产生的畸变。特别是对于广泛使用的经纬球面投影,其两极区域的扭曲失真特别严重。提出了一种全新的运动补偿预测算法,其将当前像素投影回球面并寻找其匹配的参考像素块。在此基础上利用球面上运动相对一致,提出了一种新的运动向量预测(MVP)模式。实验结果表明:该算法与传统高效视频编码器(HEVC)相比,最高提升3.65%平均码流差(BDBR)的编码性能。

[关键词] 全景视频;虚拟现实;运动补偿;视频编码

[Abstract] Virtual reality can provide immersive panoramic video for the end user experience by image rendering technology. Different from the traditional 2D video, 360°panoramic video includes the translational motion and the affine motion. The affine motion is mainly due to the distortion generated by the spherical mapping. For the latitude and longitude spherical projection which is widely used, the polar area distortion is especially serious. In this paper, a novel motion compensated prediction scheme is proposed to deal with non-translational motions for panoramic video. Furthermore, a compensation of motion vector predictor (MVP) is employed on the sphere plane to further improve the coding performance. Experimental results show that the proposed method reaches a maximum 3.65% Bjontegaard Delta Bitrate (BDBR) reductions compared to high efficiency video coding (HEVC).

[Keywords] panoramic video; virtual reality; motion compensation; video coding

下载阅览: PDF