Enlacement and Interlacement Shape Descriptors

Abstract

We propose a novel approach to characterize complex 2D shapes based on enlacement and interlacement directional spatial relations. This new relational concept allows to assess in a polar space how the concave parts of objects are intertwined following a set of directions. In addition, such a spatial relationship has an interesting behavior considering the common properties in pattern recognition such as translation, rotation, scale and symmetry. A shape descriptor is defined by considering the enlacement of its own shape and the disk area that surrounds it. An experimental study carried out on two datasets of binary shapes highlights the discriminating ability of these new shape descriptors.