English

Localized Fourier Analysis for Graph Signal Processing

Signal Processing 2021-12-02 v3 Functional Analysis

Abstract

We propose a new point of view in the study of Fourier analysis on graphs, taking advantage of localization in the Fourier domain. For a signal ff on vertices of a weighted graph G\mathcal{G} with Laplacian matrix L\mathcal{L}, standard Fourier analysis of ff relies on the study of functions g(L)fg(\mathcal{L})f for some filters gg on ILI_\mathcal{L}, the smallest interval containing the Laplacian spectrum sp(L)IL{\mathrm sp}(\mathcal{L}) \subset I_\mathcal{L}. We show that for carefully chosen partitions IL=1kKIkI_\mathcal{L} = \sqcup_{1\leq k\leq K} I_k (IkILI_k \subset I_\mathcal{L}), there are many advantages in understanding the collection (g(LIk)f)1kK(g(\mathcal{L}_{I_k})f)_{1\leq k\leq K} instead of g(L)fg(\mathcal{L})f directly, where LI\mathcal{L}_I is the projected matrix PI(L)LP_I(\mathcal{L})\mathcal{L}. First, the partition provides a convenient modelling for the study of theoretical properties of Fourier analysis and allows for new results in graph signal analysis (\emph{e.g.} noise level estimation, Fourier support approximation). We extend the study of spectral graph wavelets to wavelets localized in the Fourier domain, called LocLets, and we show that well-known frames can be written in terms of LocLets. From a practical perspective, we highlight the interest of the proposed localized Fourier analysis through many experiments that show significant improvements in two different tasks on large graphs, noise level estimation and signal denoising. Moreover, efficient strategies permit to compute sequence (g(LIk)f)1kK(g(\mathcal{L}_{I_k})f)_{1\leq k\leq K} with the same time complexity as for the computation of g(L)fg(\mathcal{L})f.

Keywords

Cite

@article{arxiv.1906.04529,
  title  = {Localized Fourier Analysis for Graph Signal Processing},
  author = {Basile de Loynes and Fabien Navarro and Baptiste Olivier},
  journal= {arXiv preprint arXiv:1906.04529},
  year   = {2021}
}
R2 v1 2026-06-23T09:50:02.917Z