Object Description and Decomposition by Symmetry Hierarchies

Abstract

Symmetry is an important feature of visual scene exploration and interpretation. Similarly, hierarchical structures figure an important aspect of symmetry. Visual symmetries describe image regions that might naturally overlap or enclose smaller symmetries. For this reason, objects and scenes can be described in their overall shape as also in their decomposition into more detailed subordinate structures by symmetry hierarchies. Most hierarchical approaches in this area are based on structural, multi-scalar or multi-resolution hierarchies. In this paper, we propose a symmetry-oriented hierarchy with a related, but more cognitive meaning by describing a hierarchy of symmetry itself based on a range-based symmetry detector. We motivate and present an approach for symmetry hierarchy representation and show experiments towards object description and decomposition by symmetry hierarchies.