Retrieving 3D shapes based on their appearance

  • Authors:

  • Ryutarou Ohbuchi;Masatoshi Nakazawa;Tsuyoshi Takei

  • Affiliations:

  • University of Yamanashi, 4-3-11 Takeda, Kofu-shi, Yamanashi-ken, Japan;University of Yamanashi, 4-3-11 Takeda, Kofu-shi, Yamanashi-ken, Japan;University of Yamanashi, 4-3-11 Takeda, Kofu-shi, Yamanashi-ken, Japan

  • Venue:
  • MIR '03 Proceedings of the 5th ACM SIGMM international workshop on Multimedia information retrieval
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, we propose an algorithm for shape-similarity comparison and retrieval of 3D shapes defined as polygon soup. One of the issues in comparing 3D shapes is the diversity of shape representations used to represent these "3D" shapes. While a solid model is well-defined and is easier to handle, others such as polygon soup poses many problems. In fact, a polygon soup 3D model most often does not define a 3D shape, but merely an illusion of "3D shape-ness" by its collection of independent polygons, lines, and manifold meshes. The most significant feature of our 3D shape similarity comparison method is that it accepts polygon soup and other ill-defined 3D models. Our approach is to use the rendered appearance only of the model as the basis for shape similarity comparison. Our method removes scale and positional degrees-of-freedom by using normalization, and the three rotational degrees of freedom by using a combination of discrete sampling of solid angles and a rotation-invariant 2D image similarity comparison algorithm. Evaluation experiments showed that, despite its simplicity, our approach worked quite well despite its simplicity.