全景视频基于纬度分段映射算法

在全景视频映射过程中,局部区域往往存在严重过采样或欠采样的问题。针对该问题,提出一种基于纬度分段映射算法(Latitude-Based Segmentation Mapping,LBSM)。用4条纬线将包含全景的球面切分为5个部分,并分别对这5部分用不同的方式进行采样;将所有采样后的像素排布为一整个紧凑的矩形以便于编码;通过扩边方法消除编码过程中产生的痕迹,以此提升视频的主观质量。实验表明,与经纬图映射算法(ERM)、双环带映射算法(DRM)、二十面体映射算法(ISM)相比,LBSM算法能够节省更多码率。与基准算法ERM相比,在使用SPSNR和WSPSNR代替PSNR的情况下,码率分别平均降低11.39%和11.76%,压缩效果显著提升。

LATITUDE-BASED SEGMENTATION MAPPING FOR PANORAMIC VIDEO

In order to solve the problem that serious oversampling or undersampling occurs in the process of panoramic video mapping,this paper proposed a latitude-based segmentation mapping(LBSM).We divided the sphere containing panoramic views into 5 parts by 4 latitude lines to sample the 5 parts with different sampling methods respectively,and arranged all pixels into a compact rectangle for encoding.We used the padding method to eliminate the tracks generated in the encoding process to improve the subjective quality of video.Experimental results demonstrate that the LBSM scheme saves more bitrate compared with equirectanglar mapping scheme(ERM),double-ring mapping scheme(DRM) and icosahedral mapping scheme(ISM).In the case of using SPSNR and WSPSNR,the LBSM scheme reduces the bitrate by 11.39% and 11.76% respectively compared with the ERM scheme.