针对可视化流动图像边缘扩散及噪声和异常点影响使得光流计算稳健性较差的问题, 提出一种基于物理学的稳健光流计算方法以改善光流计算稳健性。所提算法在基于物理学光流方法中引入各向异性滤波器以增强边缘光流稳健性, 并增加惩罚因子以减少噪声及异常点对光流计算的影响, 而后基于变分方法极小化光流能量函数以求解欧拉-拉格朗日方程, 最后通过迭代方法求得速度场。仿真结果表明, 与传统的Lucas-Kanade,Horn-Schunck,金字塔Lucas-Kanade以及基于物理的光流计算方法相比, 所提算法可显著减少边缘和角落区域光流扩散, 改善针对噪声及异常点的稳健性, 从而得到具有较好稳健性的速度场。

Focusing on the issue that the computation of the optical flow has a weak robustness to the edge diffusion, noise and outliers of the visible flow image, an improved approach of physics-based optical flow computation is developed to enhance the robustness of the optical flow computation. The proposed algorithm introduces the anisotropic filter into the method of the physics-based optical flow to improve the robustness of edge optical flow, and integrates the penalty factor into this approach to reduce the effect of noise and outliers on the optical flow computation. Then, the energy function of optical flow is minimized by using variation principle to solve the Euler Lagrange equation. Finally, the velocity field is obtained by exploiting the iterative method.Simulation results show that, compared with the traditional Lucas-Kande, Horn-Schunck, Pyramid Lucas-Kande, and the physics-based optical flow computational methods, the proposed algorithm can significantly reduce the diffusion of optical flow on edges and in corners, improve the robustness to noise and outliers, and thus obtain a velocity field with higher robustness.