基于运动估计优化的HS光流算法研究

HS(HornSchunck)光流算法检测运动物体信息的计算量较大,不能检测到平滑区域的光流信息,并且在复杂环境下容易受到噪声影响.本文提出了一种基于运动估计的优化HS光流算法.该算法通过检测图像中的HARRIS角点,结合基于宏块的运动估计算法确定感兴趣区域,并将此感兴趣区域作为HS计算的初始运动向量,以得到光流信息.最后进行滤波去除背景噪声.试验结果表明,该算法不仅提高了计算速度,避免了背景噪声的干扰,提高了HS的鲁棒性,也解决了HS算法对平滑部分光流信息的无法检测问题.

Optical flow algorithm based on optimized motion estimation

Horn Schunch(HS) optical flow algorithm can accurately detect the movement of objects, including a complex motion. However, its computation complexity is very high and it cannot detect the optical flow of smooth region in movement. To solve these problems a new optical flow based on the motion estimation is proposed, even it is susceptible to noise in a complex environment. It computes initial vectors for HS by motion estimation and harris information, and then filters the vectors of HS. The experiment shows that the algorithm proposed can accurately detect the vectors of smooth interesting area, decrease compute times, and enhance robust of HS in a complex environment.