基于维纳线性预测的Horn-Schunck光流运动矢量优化算法

A Horn-Schunck optical flow motion vector optimization algorithm based on Wiener linear predication

针对Horn-Schunck光流运动估计的矢量中可能出现局部错误估计点的问题,提出一种基于维纳线性预测的光流运动矢量优化算法。首先,将光流运动矢量从笛卡尔坐标转换到极坐标下;其次,根据制定的判决规则,对于判决中的可疑点做进一步判定,而对于判决中的错估点采用维纳线性预测的方法进行重新估值;最后,将极坐标下的光流矢量转换到笛卡尔坐标下,到此就完成了光流运动矢量的优化。与直接求出的Horn-Schunck光流矢量相比,优化后的Horn-Schunck光流矢量中幅度和角度错估点的误差明显减小,光流矢量的准确性获得了一定程度的提高。将直接求出的Horn-Schunck光流矢量和优化后光流矢量分别应用到图像和视频序列的运动补偿中。结果表明:基于维纳线性预测的Horn-Schunck光流运动矢量优化算法取得了比较好的效果。

Aiming at the problem of local error estimate points which occurs in Horn-Schunck optical flow motion vector, we propose an optical flow motion vector optimization algorithm based on Wiener linear prediction. Firstly, optical flow motion vector is transferred from the Cartesian coordinates to the polar coordinates. Secondly, according to the rules specified in the judgment, suspicious points go through further judgment, and the miscalculated points are revalued using the Wiener linear prediction method. Lastly,the optical flow motion vector is transferred from the polar coordinates back to the Cartesian co- ordinates, and the optimization of optical flow motion vector is accomplished. The deviation of magnitude and angle misjudged points in optimized Horn-Schunck optical flow motion vector is decreased obviously in comparison with that of the original Horn-Schunck optical flow motion vector method. Experimental results demonstrate that the proposed algorithm enhances the accuracy in motion estimation of images and video sequences to a certain extent.