Abstract
Multiscale transforms designed to process analog and discrete-time signals and images cannot be directly applied to analyze high-dimensional data residing on the vertices of a weighted graph, as they do not capture the intrinsic geometric structure of the underlying graph data domain. In this paper, we present a modular framework to adapt multiscale transforms such as the Laplacian pyramid, critically sampled filter bank, and lifting scheme so that they can be used to analyze high-dimensional data residing on the vertices of a weighted graph. Our approach is to study existing methods and develop new methods for the three fundamental operations of filtering for signals on graphs, graph downsampling, and graph reduction. Equipped with appropriate notions of these operations, we leverage the basic multiscale constructs and intuitions from classical signal processing to generate specific examples of new multiscale graph signal processing transforms.

This publication has 23 references indexed in Scilit: