Divide-and-Conquer Subspace Segmentation

Several important computer vision tasks have recently been formulated as lowrank problems, with the Low-Rank Representation method (LRR) being one recent and prominent formulation. Although the method is framed as a convex program, available solutions to this program are inherently sequential and costly, thus limiting its scalability. In this work, we explore the effectiveness of a recently introduced divide-and-conquer framework, entitled DFC, in the context of LRR. We introduce the DFC-LRR algorithm as a scalable solution to the subspace segmentation problem, presenting results that illustrate the scalability and accuracy of DFC relative to LRR. We further present a detailed theoretical analysis that shows that the recovery guarantees of DFC-LRR are comparable to those of LRR.