Abstract (EN)

In this paper, we present a new method for the segmentation of moving objects. We use one of the most powerful variational method for computing the optical ow and we exploit this information in the segmentation. This segmentation lies on well- known techniques of active contours. Since we can distinguish moving objects from static elements of a scene by analyzing norm of the optical ow vectors, this one is incorporated in a region-based active contour model in order to attract the evolving contour to moving objects contours. We also take gray level into account since it is known that optical ow information does not give the exact contours of the objects. The results provided are promising and extend the numerous works on this subject.