3D Object Retrieval Using Compact Shape Descriptor

Abstract

The need for effective and efficient 3D object retrieval approaches is emerging in a broad range of applications in science and engineering. With the increasing understanding of shape geometry and topology in the context of shape similarity, workable solutions for 3D object retrieval are being produced. In this paper, we present a novel technique for 3D Object Retrieval. The key idea of the proposed approach is based on the synergy between Heat Kernel Signatures (HKS) and Bag of Features (BoF) paradigm. For a given 3D model, the proposed approach considers a set of feature points, defined by an innovative feature point detection algorithm, and associate them with a compact HKS feature descriptor. Then, a vocabulary is constructed and BoF vector describing each 3D model is computed. Finally, the challenging problem of matching two given 3D models sums up to measuring the distance between their corresponding BoF distributions. The proposed approach is not only computationally efficient but also highly discriminative. It achieves state of the art results on SHREC 2011 dataset of non-rigid models, confirming its invariance to different kinds of deformation and possible noise.