A 3D Shape Descriptor Based on Hadamard Transform and Spherical Harmonic Transform

Shape matching plays an important role in many application fields. In this paper, we propose a novel rotation invariant 3D shape descriptor based on Hadamard transform and spherical harmonic transform. In our method, a 3D model is represented as a collection of spherical functions to preserve shape information as much as possible, and the shape similarity is directly defined by the difference of the character functions of 3D models. Retrieval experiments show that our method performs better than many other existing 3D matching methods based on spherical harmonics.

• Type: Conference
• Year: 2007
• Where: MVA
• Authors: Guyu Xin, Hongbin Zha
• Keywords: Computer Vision, Invariant 3D Shape, MVA 2007, Shape Matching, Spherical Harmonic Transform