COMPUTING OCCLUDING AND TRANSPARENT MOTIONS

Computing the motions of several moving objects in image sequences inv olves simultaneous motion analysis and segmentation. This task can bec ome complicated when image motion changes significantly between frames , as with camera vibrations. Such vibrations make tracking in longer s equences harder, as temporal motion constancy cannot be assumed. The p roblem becomes even more difficult in the case of transparent motions. A method is presented for detecting and tracking occluding and transp arent moving objects, which uses temporal integration without assuming motion constancy. Each new frame in the sequence is compared to a dyn amic internal representation image of the tracked object. The internal representation image is constructed by temporally integrating frames after registration based on the motion computation. The temporal integ ration maintains sharpness of the tracked object, while blurring objec ts that have other motions. Comparing new frames to the internal repre sentation image causes the motion analysis algorithm to continue track ing the same object in subsequent frames, and to improve the segmentat ion.