摘要
提出了一种基于帧矢量叠加和自适应中止技术的快速运动估计方法。该方法在不改变搜索策略的前提下,通过采用帧矢量叠加技术,对运动矢量进行传递,等效扩大了搜索区域,显著减少了相邻帧之间对应像素最小绝对误差值的跳变,运算量减少至21.8%;引入中止判决准则,在匹配运算中,自适应的改变决定匹配次数的阈值,将运算量减少为40.2%。以全搜索块匹配算法为例的实验表明,该算法在保持精度不变的前提下,运算量最终减少到8.72%。
A method of fast motion estimation based on frame vector accumulative carry and adaptive break algorithms is proposed. With the frame vector-addition technology, the motion vector is transferred and the search region is broadened equivalently. The change of MAD between adjacent frames is reduced significantly and the calculation load is lowered to 21.8% while the search method is not changed. Also with the adoption of the principle of adaptive break, the threshold to decide the number of times of block matching is adaptive and the average number of times is reduced to 40.2% in processing of the experimental images, Experimental data on full search block matching algorithm indicates that the proposed method keeps the accuracy while the computation load is reduced to 8.72%.
出处
《电视技术》
北大核心
2006年第4期7-10,共4页
Video Engineering
基金
中国航天支撑基金(HG30
03)资助
关键词
电子稳像
运动估计
运动矢量
electronic image stabilization
motion estimation
motion vector
