摘要
EPZS是联合多视点视频编码(JMVC,Joint Multi-view Video Coding)运动估计中采用的一种预测搜索算法,其搜索速度慢.针对EPZS算法的性能不足,我们在预测矢量集合、搜索模型、阈值设置和搜索策略四个方面进行改进,提出了一种混合快速搜索算法.在联合多视点视频编码测试平台JMVC8.3中,对三个由平行摄像机采集的多视点视频测试序列BallRoom、Exit和Vassar进行测试.实验结果表明:在保证视频重建质量和码率的前提下,与Jmvc中的EPZS算法相比,编码速度平均提高了55.66%~69.62%,改进算法的效果明显,编码效率得以提高.
EPZSI Enhanced Predictive Zonal Search) is a fast hybrid motion vector search algorithm adopted by the motion estimation of Joint multiview video coding. EPZS algorithm is a predictive search algorithm but its search velocity is slow. For EPZS algorithm' s disadvantages,a fast hybrid search algorithm based on multiview video coding is proposed. The optimization is suggested mainly from the following four aspects: predictive vector sets, search model, threshold setting as well as search strategy. Experimental results on JMVC8.3 show that the performance of modified algorithm is so good that coding efficiency is also improved, on the promise of right code rate and reconstruction video quality comparing to EPZS in JMVC. The average encoding speed increased from 55.66 to 69.62 percent.
出处
《小型微型计算机系统》
CSCD
北大核心
2013年第5期1176-1180,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(60972037)资助
省部产学研项目(2012B091100495,2009B090300267)资助
深圳市基础研究计划项目(JCYJ20120613113419607)资助
校重点项目(2213k3190007)资助
关键词
运动估计
多视点视频编码
混合搜索
EPZS
编码效率
motion estimation
MVC
hybrid search
enhanced predictive zonal search
coding efficiency