电子稳像中运动矢量一致性的优化算法

An algorithm to optimize coherence of motion vector in electric image stabilizing

针对电子稳像技术传统块匹配法求运动矢量的固有缺点,采取四叉树———一种变块匹配方法计算运动矢量,使求得的运动矢量有较好的一致性.与其它变块匹配方法相比,该方法克服了其它变块匹配法的不足.实验表明,使用该法求出的运动矢量有较好的一致性和准确性.

The block matching algorithm based on MAD (mean absolute difference) is extensively used in the motion estimation. In order to eliminate the inherent shortcomings of the classical block matching algorithm, quad tree--a size-variable block matching technique is proposed to optimize the coherence of motion vectors. Also this algorithm is compared with the other variable block matching methods. The experimental results show that this algorithm may produce reliable and coherent motion vectors.