In this paper, a novel approach to face clustering is proposed. The aim is the completely unsupervised extraction of planes in a polygonal a mesh, obtained from a 3D reconstruction process. In this context, 3D coordinates points are inevitably affected by error, therefore resiliency is a primal concern in the analysis. The method is based on the Mean Shift clustering paradigm, devoted to separating modes of a multimodal non-parametric density, by using a kernel-based technique. A critical parameter, the kernel bandwidth size, is here automatically detected by following a well-accepted partition stability criterion. Experimental and comparative results on synthetic and real data validate the approach.
3D Objects Face Clustering using Unsupervised Mean Shift
FUSIELLO, Andrea
2007-01-01
Abstract
In this paper, a novel approach to face clustering is proposed. The aim is the completely unsupervised extraction of planes in a polygonal a mesh, obtained from a 3D reconstruction process. In this context, 3D coordinates points are inevitably affected by error, therefore resiliency is a primal concern in the analysis. The method is based on the Mean Shift clustering paradigm, devoted to separating modes of a multimodal non-parametric density, by using a kernel-based technique. A critical parameter, the kernel bandwidth size, is here automatically detected by following a well-accepted partition stability criterion. Experimental and comparative results on synthetic and real data validate the approach.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
