A Class of Sparse Unimodular Matrices Generating Multiresolution and Sampling Analysis for Data of Any Length

  • Authors:
  • N. D. Atreas;C. Karanikas;P. Polychronidou

  • Affiliations:
  • natreas@csd.auth.gr;karanika@csd.auth.gr;ppolychr@csd.auth.gr

  • Venue:
  • SIAM Journal on Matrix Analysis and Applications
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

We introduce a class of sparse unimodular matrices $U^m$ of order $m \times m$, $m = 2,3, \ldots\,$. Each matrix $U^m$ has all entries $0$ except for a small number of entries $1$. The construction of $U^m$ is achieved by iteration, determined by the prime factorization of a positive integer $m$ and by new dilation operators and block matrix operators. The iteration above gives rise to a multiresolution analysis of the space $V_m$ of all $m$-periodic complex-valued sequences, suitable to reveal information at different scales and providing sampling formulas on the multiresolution subspaces of $V_m$. We prove that the matrices $U^m$ are invertible, and we present a recursion equation to compute the inverse matrices. Finally, we connect the transform induced by the matrix $U^m$ with the underlying natural tree structure and random walks on trees.