LAUSR

Subspace clustering methods have been widely studied recently. When the inputs are two-dimensional (2-D) data, existing subspace clustering methods usually convert them into vectors, which severely damages inherent structures and relationships from original data. In this letter, we propose a novel subspace clustering method for 2-D data. It directly uses 2-D data as inputs such that the learning of representations benefits from inherent structures and relationships of the data. It simultaneously seeks image projection and representation coefficients such that they mutually enhance each other and lead to powerful data representations. An efficient algorithm is developed to solve the proposed objective function with provable decreasing and convergence property. Extensive experimental results verify the effectiveness of the new method.