利用点线对应计算彩色时变图像光流场

A New Method of Optical Flow Estimation from Color Image Sequences Using Feature Correspondence

时变图像光流场计算是当今计算机视觉研究的前沿方向 ,彩色时变图像光流场是光流场计算从灰度图像到彩色图像的推广。利用彩色图像提供的更丰富的光学信息 ,使光流场计算的不适定问题转换成适定问题。文中首先利用图像边缘检测与 Hough变换 ,提取图像的边缘点和边缘线。然后把提取出的边缘点和边缘线分别作为图像的特征点和特征线。最后利用彩色时变图像中的特征点线对应 ,把基于特征的方法和基于光流场的方法结合起来 。

We present a new method of optical flow estimation from color image sequences using the feature correspondence. Section 1 discusses edge detection and Hough transformation. We use Hough transformation to determine the feature points on the contour. Section 2 discusses in detail how to compute optical flow of time varying color image, using the feature points on the contour. Mathematically we use Eq. (6) to compute this optical flow. Experimental results, given in Figs.2 and 4, show preliminarily that our new method provides a good estimation of the optical flow field, particularly in two respects: (1) features are well extracted; (2) continuity of optical flow field is satisfactory.