利用Gabor滤波的相位图像进行光流估计

Estimation of Optical Flow for Phase Image Using Gabor Filter Banks

提出一种联合Gabor滤波器和加权最小二乘法估计序列图像光流的方法,用Gabor滤波器的滤波结果作为新的图像特征来计算空间和时间相位梯度.为了获得较为精确的结果,用加权最小二乘法来估计序列图像的光流.实验结果表明,文中算法在光流的计算精度上要优于Lucas,Fleet等的方法.

In the paper, a new algorithm for estimating the optical flow of image sequences is presented, by combining Gabor filter banks with weighted least-squares. The results of Gabor filter banks are used as the new features of image for the computation of the spatial-temporal phase gradient. In order to obtain more accurate results, the optical flow of sequence image is estimated by the method of weighted leastsquares. The results of the experiments indicate that more accurate results of optical estimation can be obtained compared with the methods suggested by Lucas, Fleet, et al.