一种鲁棒性的光流算法研究

Research on a Robust Optical Flow Algorithm

提出了一种用于计算帧间运动速度及速度变化都较大时的鲁棒性光流算法 .通过对时间梯度误差的分析 ,得到了新的梯度约束方程 ;另一方面将各向异性扩散方程引入方向平滑约束 ,以改善包括运动不连续处的流速计算精度 .以此建立的多栅格微分光流算法不但可以有效地缓解误差传播效应 ,而且对大的运动速度和速度变化有很好的适应性 .实验结果表明 ,本文提供的算法不但能产生密度 10 0 %的光流场 ,而且对较大的运动速度及速度变化具有很好的计算精度和鲁棒性 .

A robust optical flow algorithm for larger motion velocity and lager velocity change was presented. A novel gradient constraint equation was obtained through analyzing the error of temporal gradient. On the other hand, the anisotropic diffusion equation was introduced in the oriented smoothness constraint so that the computation of flow velocity in the motion discontinuity can be improved. According to these improvements, the multi-grid differential optical flow algorithm can not only alleviate the error propagation effect, but also adapt itself to larger motion velocity and lager velocity change. The experimental results have shown that the flow field of density 100% can not only be created by the algorithm proposed in this paper, but also be provided with good computation accuracy and robustness for lager motion velocity and lager velocity change.