Robust Subspace Segmentation by Self-Representation Constrained Low-Rank Representation

Abstract

Low-rank representation (LRR) and its variants have been proved to be powerful tools for handling subspace segmentation problems. In this paper, we propose a new LRR-related algorithm, termed self-representation constrained low-rank presentation (SRLRR). SRLRR contains a self-representation constraint which is used to compel the obtained coefficient matrices can be reconstructed by themselves. An optimization algorithm for solving SRLRR problem is also proposed. Moreover, we present an alternative formulation of SRLRR so that SRLRR can be regarded as a kind of Laplacian regularized LRR. Consequently, the relationships and comparisons between SRLRR and some existing Laplacian regularized LRR-related algorithms have been discussed. Finally, subspace segmentation experiments conducted on both synthetic and real databases show that SRLRR dominates the related algorithms.