Video segmentation using active contours on a group of pictures

In this paper we address a new method for motion camera video segmentation. Most video segmentation methods use the temporal gradient between two consecutive frames and therefore are not robust to light variations and noise. Thus we present a new region based active contour segmentation on a group of pictures. This new segmentation proceeds in two steps: first a mosaic is computed to generate the background sequence of the video; then color region based active contour segmentation is applied on each video frame, related to its background. Finally, we show experimental results obtained on real sequences after each step of the algorithm i.e. the background mosaic and the final segmented objects.