Abstract: The optical flow estimation algorithm based on the optical flow constraint and smoothness constraint presented by Horn and Schunk is one of the important algorithms of the image motion estimation. The algorithm cannot acquire accuracy motion parameter estimation at low-gradient points. At the same time, the present improved methods required artificial selected parameters and when the threshold value was set too high the object area would be unperfected. Two optical flow estimation methods were presented by modifying the optical flow basic constraint weighted function. Experiment results show that the improved methods can depress the repression of reliable optical flow when the threshold value was set too high and improve the self-adaptive ability what lays a good foundation for moving object detection and tracking.

Key words: Optical flow field, Global smoothness constraint, Weighted function, Self-adaptive, Motion estimation