Shape feature decomposition of regularized objects

This paper deals with the problem of shape feature identification from boundary representations of objects belonging to the r-set domain. Existing recognition systems usually apply to two manifold objects, while in our approach we consider also regularized non-manifold objects. The system performs a shape decomposition of an object into a union of simpler subparts: two-manifold components and concave/convex facesets. Such a decomposition does not depend on the sequence of operations used to build the object, it is unique, invariant under geometric transformations and its complexity is comparable with the complexity of the original description