摘要
H.264是由ISO/IEC和ITU-T共同发布的较新的视频编解码标准,而JM模型是JVT官方发布的H.264标准测试模型。在H.264编码标准中,运动估计是最重要而最耗时的过程,对于多预测模式、多参考帧,更细化的运动矢量来说更是如此。在JM模型中,可选的运动估计方法有三种,通过对JM10.2源码中的UMHexagonS算法进行分析,改进宏块中绝对误差和计算步骤,在保证视频序列信噪比的前提下减少运动估计的耗时。实验结果表明该方法对视频序列处理耗时平均减少达5%~15%,而信噪比衰减不超过0.3%。
H.264 is a new standard for video coding which is published by both ISO/IEC and ITU-T while JM is the standard test model published by joint video team (JVT). Motion estimation is the most important and consuming part of H.264 codec, specifically with multi prediction modes, multi reference frames, and higher motion vector resolution. There are three methods for motion estimation in JM model. Based on the study of UMHexagonS algorithm in JM 10.2, a new method by improving the calculation procedure of SAD in macro block is proposed in order to reduce time consuming of motion estimation with minimal loss in video bitrates. Results show that the improved method of processing video sequences can reduce the time consuming by 5%~15% while keep a minimum decay of SNR no more than 0.3%.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2009年第S1期149-150,154,共3页
Journal of System Simulation
基金
视频图像中人体目标检测算法研究
基于MPEG4的图象压缩处理及网络传输机制研究