A new gradient-based optical flow method and its application to motion segmentation

Optical flow estimation for motion image sequences has been widely known as one of the most fundamental problems in motion image analysis. Many methods have been proposed for computing optical flow; among them, gradient-based methods are the best known and most used. In this paper, we propose a new method to compute optical flow by applying multiple gradient constraints to a set of image pixels within a given neighborhood. Compared with other commonly-used gradient-based methods, our proposed method is able to provide a more reliable estimation without reducing the optical flow density. We also implemented our new method for the motion segmentation of image sequences and obtained promising results.