一种改进的Lucas-Kanade光流估计方法

Improvement of Lucas-Kanade Method for Optical Flow Estimation

摘要:

Lucas-Kanade算法是一种重要的光流估计技术,在计算机视觉领域有广泛的应用。但对于图像纹理欠丰富的区域,Lucas-Kanade算法光流估计结果较差。文章提出了一种改进的Lucas-Kanade算法:首先,计算跟踪置信度因子；然后,通过阈值检测剔除不可靠的光流估计结果；最后,进行邻域填充。该方法改善了纹理欠丰富区域的光流估计,对矮人模型图像序列进行了实验,证明了方法的有效性及优于Poelman方法和Chiba方法。

Abstract:

The Lucas-Kanade method is one of the best optical flow estimation techniques, because it is fast and simple to implement. But a big problem of it is that many regions of typical real images do not contain sufficient texture to produce reliable optical flow estimation. An improvement of Lucas-Kanade method was proposed to remedy the problem, which used tracking confidence to threshold the estimation of any resolution level and recomputed the optical flow by their neighboring regions. Experiments using image sequences of dwarf model showed that our method was effective and better than Poelman’s and Chiba’s method.