Kernel Structure Constrained Low Rank Representation for Manifold Clustering

Because the data in many computer vision problems usually has the structure of mixing manifolds,a manifold clustering method is proposed in this paper. By designing the special iteration of 2,1 norm to overcome the technical problem, the proposed method kernelizes structure-constrained low-rank representation. Experimental results on Hopkins 155, Caltech 256, etc. confirm the effectiveness of the kernel structure constrained low-rank representation for manifold clustering.