Kernel Spectral Curvature Clustering KSCC - Computer Science > Computer Vision and Pattern RecognitionReport as inadecuate




Kernel Spectral Curvature Clustering KSCC - Computer Science > Computer Vision and Pattern Recognition - Download this document for free, or read online. Document in PDF available to download.

Abstract: Multi-manifold modeling is increasingly used in segmentation and datarepresentation tasks in computer vision and related fields. While the generalproblem, modeling data by mixtures of manifolds, is very challenging, severalapproaches exist for modeling data by mixtures of affine subspaces which isoften referred to as hybrid linear modeling. We translate some importantinstances of multi-manifold modeling to hybrid linear modeling in embeddedspaces, without explicitly performing the embedding but applying the kerneltrick. The resulting algorithm, Kernel Spectral Curvature Clustering, useskernels at two levels - both as an implicit embedding method to linearizenonflat manifolds and as a principled method to convert a multiway affinityproblem into a spectral clustering one. We demonstrate the effectiveness of themethod by comparing it with other state-of-the-art methods on both syntheticdata and a real-world problem of segmenting multiple motions from twoperspective camera views.



Author: G. Chen, S. Atev, G. Lerman

Source: https://arxiv.org/







Related documents