摘要
为了提高视频编码运动估计中运动矢量预测的速度和准确性,提出一种基于贝叶斯决策的视频自适应运动估计算法。该算法充分利用贝叶斯理论、运动矢量的空间一致性和已编码帧对当前帧的影响,根据视频中前一帧和当前帧的已搜索宏块的运动信息,以目标宏块周围3个宏块的运动矢量与目标宏块的运动矢量空间距离最小为原则,利用贝叶斯决策来得到目标宏块运动矢量的预测值。实验表明,该方法在图像重建质量基本不变的情况下,比DS,ARPS和ARPS-3具有更快的搜索速度。
An adaptive video motion estimation algorithm based on Bayesian decision is proposed to improve the speed and accuracy of the motion vector prediction in video motion estimation. Bayesian theory, space consistency of MV and the encoded frame＇s influence on current frame are made full use. Based on the known video blocks of the previous frames and the current frame, the probabilities of minimum space distance among the motion vector of the target search block and the three surrounding blocks are calculated, and the MV prediction is determined according to the Bayesian decision. Compared to the DS, the ARIAS and the ARPS-3 algorithms, experimental results show that the proposed method provides faster search speed while the reconstruction quality holds unchanged.
出处
《电视技术》
北大核心
2011年第5期25-28,共4页
Video Engineering
基金
江苏省自然科学基金项目(BK20080544)
关键词
视频编码
运动估计
自适应
贝叶斯决策
video coding
motion estimation
adaptive
Bayesian decision
作者简介
陈天壮（1986-），硕士生，主研图像处理、视频编解码；
梁久祯（1968-）。博士，教授。主研智能信息处理与模式识别、图像处理.
韩军（1965-）．博士。副教授，主研视频编解码、图像处理。
