色彩梯度恒常性的光流场估计算法

Optical flow estimation algorithm using color gradient invariance

为提高光流估计的鲁棒性,在彩色图像光流场计算中色彩恒常假定的基础上,进一步假定色彩梯度在运动中保持不变,据此提出了一种基于色彩梯度恒常性假设的光流求解方法,以色彩梯度构成光流基本方程,并对其施加全局平滑约束,以Gauss-Seidel迭代求解光流场,并用中值滤波去除光流场中的异常分量.实验表明,该方法相对于灰度图像序列及彩色图像序列的经典光流场估计算法可取得更好的估计效果.

In optical flow estimates, it is usually necessary to assume that the fluid being measured has an unvarying color. For the sake of improving the robustness of optical flow estimates, an additional assumption is adopted that the color gradient of the measured substance is also uniform. Based on these two assumptions, a new optical flow estimation method is presented in this paper, in which the basic optical flow equation is derived in terms of the color gradient, with global smooth constraints applied on it. The flow velocity is solved through Gauss-Seidel iteration and the abnormal velocity component is removed by median filtering. The results of experiments show that this method yields better estimating results than the classic gray image sequence method and the color image sequence method.