摘要
H.264标准中的运动估计部分占了50%以上的运算开销.在分析了图像运动矢量的中心偏置特性的基础上提出了基于十字模板的梯度快速搜索算法(CGS).该方法提取搜索中两个点蕴含的矢量梯度,配合十字模板,能够快速确定搜索方向和范围,逼近最优匹配点.运动搜索点数比六边形算法降低了36%以上,图像PSNR值比全搜索只下降了0.01dB.通过实验分析也证明了运动矢量中心偏置特性,该算法具有一定的应用价值.
Motion estimation costs over 50 % caculation in H. 264 standard. A new fast search algorithm-Cross model Grads Search algorithm（CGS） is proposed based on the center- based characteristics of motion vector probability distribution. By looking for the motion vector grads in two points and using the cross model, the algorithm can quickly establish the search direction and range, and aproach to the best suited points. Its mo- tion search points decrease by more than 36 % compared with the points of Hexagonal algorithm, and the PSNR value only decreases by 0.01dB compared with Full Search algorithm. Experimental results show that the center-based characteristic of motion vectors and CGS is valuable in applieations.
出处
《四川大学学报（自然科学版）》
CAS
CSCD
北大核心
2008年第4期804-808,共5页
Journal of Sichuan University(Natural Science Edition)
关键词
H
.264
运动估计
中心偏置
十字模板的梯度搜索
率失真代价
H. 264, motion estimation, center-based characteristic, cross model grads search algorithm, ratedistortion cost
