Smoothness of Multiple Refinable Functions and Multiple Wavelets

Rong-Qing Jia, Sherman D. Riemenschneider, Ding-Xuan ZhouSIAM J. Matrix Anal. Appl 1999 · DOI

TL;DR

This paper introduces a novel method to precisely characterize the smoothness of multiple refinable functions and wavelets. By leveraging the p-norm joint spectral radius of matrices derived from the refinement mask, the authors provide a powerful tool for analyzing regularity, which is crucial for applications in signal processing and approximation theory.

What Problem Does This Paper Address?

In fields like signal processing and numerical analysis, understanding the "smoothness" of functions is paramount. Refinable functions and wavelets, which are fundamental building blocks for these applications, are often defined by systems of refinement equations. The smoothness of these functions dictates their approximation power and performance in various applications. While previous work has addressed this for single functions, analyzing the smoothness of multiple refinable functions and wavelets simultaneously presents a significant challenge. Existing methods often rely on factorization techniques that can become computationally complex and may not generalize well. This paper aims to bridge this gap by developing a more direct and computationally efficient approach to characterize the smoothness of these multi-dimensional systems.

Methods

The core of the paper's methodology lies in relating the smoothness of refinable functions to the spectral properties of matrices derived from the refinement mask. Specifically, the authors introduce the concept of the p-norm joint spectral radius of a collection of matrices. They show that the optimal smoothness of a refinable function vector can be bounded from below by an expression involving this joint spectral radius.

Key techniques include:

  • Fourier Transforms: Used to transform the refinement equations into a more manageable form.
  • Matrix Theory: The analysis heavily relies on properties of matrices, particularly their spectral radii and invariant subspaces.
  • Generalized Lipschitz Spaces: These spaces are used to formally define and measure the smoothness of functions.
  • Block Matrix Operators: The refinement mask is represented by bi-infinite block matrices, and their behavior on specific invariant subspaces is analyzed.
  • Iterative Relations: The paper derives and utilizes iterative relations for the refinement mask coefficients, which are crucial for connecting the mask to the spectral properties.

A significant contribution is the reduction of matrix sizes by restricting operators to minimal common invariant subspaces, making the computational implementation more tractable.

Datasets

Not discussed in the paper. The paper focuses on theoretical analysis and provides examples with specific masks, rather than using datasets for empirical evaluation.

Key Insights and Ideas

  • Spectral Radius as a Smoothness Indicator: The paper establishes a direct link between the p-norm joint spectral radius of matrices derived from the refinement mask and the smoothness of the corresponding refinable functions. This is a powerful theoretical insight.
  • Invariant Subspaces Simplify Analysis: By identifying and utilizing minimal common invariant subspaces, the authors significantly simplify the analysis of large block matrices, making the approach computationally feasible.
  • Unified Framework for Multiple Refinable Functions: The method provides a general framework applicable to systems of multiple refinable functions and wavelets, overcoming limitations of previous approaches.
  • Explicit Smoothness Characterization: The paper demonstrates that for certain cases, the optimal smoothness can be explicitly calculated, offering precise quantitative results.

Key Results

The paper's main result is a theorem that provides a lower bound for the optimal smoothness (critical exponent) of a vector of compactly supported refinable functions. This bound is given by:

$v_p(\phi) \geq 1/p - \log_2 \rho_p(A_0|_V, A_1|_V)$

where $\rho_p(A_0|_V, A_1|_V)$ is the p-norm joint spectral radius of the operators $A_0$ and $A_1$ restricted to a specific invariant subspace $V$. The paper also shows that under certain conditions (stability of shifts and a condition on $k$), this inequality becomes an equality, providing an exact characterization.

The paper also presents detailed analyses of two specific examples, explicitly calculating the smoothness for various $p$ values. For instance, in Example 4.1, they find $v_p(\phi) = 1 + 1/p$ or $v_p(\phi) = -\log_2|s|$ depending on the parameter $s$.

Limitations and Future Work

  • Stability Condition: The equality in the main theorem holds when the shifts of the refinable functions are stable. Checking this stability condition for arbitrary masks can be a separate challenge.
  • Computational Complexity for Large Systems: While the invariant subspace approach reduces complexity, analyzing very large systems of refinable functions might still be computationally intensive.
  • Focus on Lp Spaces: The primary focus is on $L_p$ spaces. Extending the analysis to other function spaces or distributions could be a future direction.

The authors suggest that their methods can be applied to multiple wavelets of several variables, indicating a potential avenue for future research. The detailed examples also demonstrate the applicability of their approach to constructing and analyzing specific families of wavelets.

Summary updated

BibTeX
@article{jia1999smoothness,
  author = {Rong-Qing Jia and Sherman D. Riemenschneider and Ding-Xuan Zhou},
  title = {{Smoothness of Multiple Refinable Functions and Multiple Wavelets}},
  journal = {SIAM J. Matrix Anal. Appl},
  year = {1999},
  doi = {10.1137/s089547989732383x},
}
Figure