Logo image
Statistical spherical wavelet moments for Content-based 3D model retrieval
Conference paper   Open access

Statistical spherical wavelet moments for Content-based 3D model retrieval

H. Laga and M. Nakajima
25th Computer Graphics International Conference (CGI) 2007 (Petropolis, Brazil, 30/05/2007–02/06/2007)
2007
pdf
spherical wavelet.pdfDownloadView
Open Access

Abstract

The description of 3D shapes with features that are invariant under similarity transformations is one of the challenging issues in content-based 3D model retrieval. In this paper we show that shape sampling affects significantly the rotation invariance of existing shape descriptors. Then we propose a new parameterization method that samples uniformly the shape which is then fed to a spherical wavelet analyzer to extract discriminative features. We introduce new shape descriptors based on higher order statistical moments of the sphericalwavelet sub-bands of the spherical shape function. The proposed descriptors are compact and invariant under similarity transformations.We demonstrate their efficiency, using the Princeton Shape Benchmark, regarding the computational aspects and retrieval performance.

Details

Metrics

69 File views/ downloads
45 Record Views
Logo image