HIERARCHICAL SEGMENTATION-BASED APPROACH TO MOTION ANALYSIS

This paper presents an algorithm for motion analysis that has been use d for identifying moving targets in a sequence of images. The algorith m is based on first generating a hierarchical segmentation of each ima ge and then identifying corresponding regions from image to image. Usi ng hierarchical image descriptions allows regions to be matched from f rame to frame even if scene changes cause the regions to appear at dif ferent scales in the segmented image. Hierarchical segmentation also a llows the algorithm to isolate the stable regions in each image which are most effective for image registration and to focus computation on the range of resolutions that are relevant for a particular applicatio n. Performing registration at the region level has severtal important advantages over traditional optical flow or feature point matching alg orithms. These advantages include better performance in the presence o f scene changes and noise and the capability for identifying successfu lly large motions between frames. Following segment registration, the algorithm estimates the motion of the background and of candidate targ ets. Moving targets are identified based on their motion relative to t he background. The effectiveness of the algorithm is demonstrated by e xperiments on image sequences containing large motions.