Subspace Clustering via Graph Regularized Sparse Coding

Sparse coding has gained popularity and interest due to the benefits of dealing with sparse data, mainly space and time efficiencies. It presents itself as an optimization problem with penalties to ensure sparsity. While this approach has been studied in the literature, it has rarely been explored within the confines of clustering data. It is our belief that graph-regularized sparse coding can lead to a data space with discriminative features that can be clustered so that these clusters can be applied to various areas of interest in the field, such as video segmentation and tracking. In this paper, we propose a framework for clustering via graph-regularized sparse coding and apply this framework to video segmentation and image patch correspondence.