Implantation temps réel d'un algorithme de segmentation par modélisation de l'architecture du système visuel biologique

At the present time, realtime image segmentation provides us many problems concerned its implementation and its achievements. A good alternative to the classical segmentation methods are those deduced from biological visual systems but they are very long to perform on Von Neuman architecture. Our algorithm uses a biological visual modelisation to perform an edge detection. We evaluate their characteristics as for instance computing time, memory or communication costs at every stage of processing. A scheduling of the main tasks is designed in a multiprocessor system. The obtained calculator would be useful to on-board systems that need efficient and robust algorithms for a large range of images and that without parameters settings.