MULTISTAGE SEGMENTATION OF OPTICAL-FLOW FIELD

This paper presents object-based segmentation of optical flow field an d its motion estimation method. A major difficulty in estimating gener al motion is that it requires a large area of support in order to achi eve a good estimation. Unfortunately, when the supporting area is larg e, it is very likely to have multiple moving objects. Thus, object-bas ed segmentation and general motion estimation are interdependent probl ems. To solve the-problem, we propose a multi-stage segmentation metho d which groups progressively the flow field into segments according to a hierarchy of motion models. The basic idea is to group uniform subr egions with respect to a simple motion model Into uniform regions with respect to a complex model using region growing technique. By the met hod, we can segment the flow field into several parabolic patches. For head-and-shoulder images, simulation results show that it can segment the flow held into meaningful regions such as head and shoulder.