Low-Complexity Reconstruction of Sampled Graph Signals in Local Graph Fourier Subspaces

In our prior work, we have introduced local graph Fourier frames (LGFFs) as a flexible and powerful modeling and analysis tool for graph signals. The most important practical advantage of LGFFs is their outstanding computational efficiency. In this paper, we discuss the vertex-domain sampling and interpolation (recovery) of graph signals that live in an LGFF subspace. We formulate perfect reconstruction conditions and develop low-complexity recovery algorithms that also work in the presence of measurement noise. Furthermore, we discuss interesting special cases and illustrate our framework by numerical experiments.