Representing the Surface of Objects by Combinatorial Pyramids

This paper introduces a new method to rep- resent the surface of objects using two dimensional com- binatorial maps. The classical de nition of two dimen- sional combinatorial maps is extended here by adding a \back face" that corresponds to the non{visible part of the object. As a rst step, every object in the scene is extracted in one image pyramid, where levels are re- lated by means of coordinates. Finally, an algorithm to complete the description of the surface is presented. Such representation is translation, rotation and scale invariant. It will allow to update information about movements and parts of the object that become visible, reducing the complexity and the computational time.