Alternating direction method for sparse subspace clustering

Subspace clustering has important and wide applications in computer vision and pattern recognition. Sparse subspace clustering constructs a sparse similarity graph for spectral clustering by using ℓ₁-minimization based coefficients, and provide an efficient method for clustering data belonging to a few low-dimensional linear subspaces. An alternating direction method is proposed to deal with noise by modifying the sparse optimization program to incorporate the corruption model. The method does not require initialization and it is computationally efficient. Motion segmentation experimental results show that the proposed method performs better than the competitive state-of-the-art subspace clustering methods.