Sampling of graph signals for graph-based view synthesis and image completion

Introduction We tackle the problem of sampling signals on graphs in Paley-Wiener spaces [1]. By introducing a new sampling algorithm, using the theoretical minimum number of samples, we can reconstruct perfectly the signal. In order to reduce the computational complexity of our algorithm we propose a local sampling approach, by partitioning the original graph in subgraphs and applying our sampling algorithm to each individual subgraph. We examine which graph partitioning method works best and in what cases.