Filtering sparse data with 3D tensorial structuring elements

Abstract

We address the problem of filtering three-dimensional sparse data representing real objects. The main application is to eliminate points that are not structured on surfaces. Points classified as organized can be the input for other processes. An accumulation process infers the organization of each input element. The tensorial fields used in our method act as three-dimensional structuring elements. They define normal orientations in space indicating possible surface continuations.