3D Segmentation by Spectral Clustering Using Shape Index
Rajaallah, E.; Herouane, O.; Moumoun, L.; Gadi, T.

Abstract
The 3D segmentation is a required pretreatment for many applications of 3D processing objects (Search by content, compression, animation, etc…). The design of a segmentation algorithm must take into account the purpose of the algorithm and the application domain to which it is intended. We propose in this paper a new approach to construct the recursive spectral clustering technique using a similarity matrix. This approach is based on the shape index values of adjacent faces and a hierarchical segmentation by optimal recursive cuts. The experimental results obtained are very encouraging and have allowed us to evidence the effectiveness of our approach for segmentation of 3D objects.

Full Text: PDF     DOI: 10.15640/jcsit.v2n3-4a2