Zeros of a Mask Function and the Orthogonality of the Related Scaling Function

Xingwei Zhou, Weifeng SuApplied and Computational Harmonic Analysis 2000 · DOI

TL;DR

This paper investigates the conditions under which a specific type of mathematical function, called a scaling function, exhibits orthogonality. The authors establish a precise criterion for orthogonality based on the zeros of a related "mask function," providing a complete answer to a long-standing question in the field of wavelet analysis. This work offers a deeper understanding of the relationship between the spectral properties of mask functions and the desirable orthogonality of scaling functions.

What Problem Does This Paper Address?

In wavelet analysis, scaling functions are fundamental building blocks used to construct wavelets. A crucial property for many applications of scaling functions is orthogonality, which simplifies their analysis and application. The orthogonality of a scaling function is intimately linked to the properties of a related mathematical object called a "mask function," which is a trigonometric polynomial.

Previous research had established that if a mask function has no zeros within certain intervals (e.g., [-1/4, 1/4] or [-1/2, 1/2]), then the corresponding scaling function is orthogonal. However, a key question remained: can we determine orthogonality by checking for zeros in arbitrarily small intervals? Specifically, for any small interval (0, Δ), can we identify a finite set of points such that if the mask function has no zeros at these points, the scaling function is guaranteed to be orthogonal? This paper aims to provide a definitive answer to this question.

Methods

The paper delves into the mathematical properties of mask functions and scaling functions, particularly focusing on the concept of "cycles." A cycle is a finite set of points that remains invariant under a specific transformation related to the mask function's frequency domain. The authors leverage existing results that connect the orthogonality of scaling functions to the absence of certain types of cycles where the mask function's magnitude is equal to 1.

Key to their approach are several lemmas that establish relationships between these cycles, sign sequences (sequences of +1 and -1), and numerical quantities derived from the mask function. They introduce specific sequences, denoted as $D_k$, and analyze their properties, particularly their "μ values." These μ values are shown to be related to the magnitude of the mask function at specific frequencies. The core of their method involves comparing these μ values to the interval Δ and relating them back to the existence of problematic cycles.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Cycles as Obstacles: The paper reinforces the idea that specific "conjugate cycles" in the frequency domain of the mask function are the primary obstacles to orthogonality. If the mask function is zero at all points within such a cycle, orthogonality is lost.
  • Quantifying the "Safe Zone": The authors introduce a sequence of values, $\mu(D_k)$, which effectively define thresholds. If the interval Δ is larger than these thresholds, there's a finite number of problematic cycles to check. However, if Δ is smaller than the smallest of these thresholds ($\Delta_0$), there are infinitely many such cycles, making a finite check impossible.
  • A Complete Criterion: The paper provides a complete answer to the original question. It precisely defines the threshold $\Delta_0$ and establishes that for $\Delta > \Delta_0$, orthogonality can be determined by checking a finite set of points. For $\Delta \le \Delta_0$, this is not possible.
  • Connection to Existing Results: The paper shows how previous results (like those by Cohen and Ryan, and Shann and Yen) are special cases of their more general findings, particularly when $\Delta$ corresponds to specific $\mu(D_k)$ values.

Key Results

The main result is Theorem 3.1, which provides a complete characterization of when a scaling function is orthogonal based on the zeros of its mask function within an interval $[-\Delta, \Delta]$.

  • If $\Delta \ge \Delta_0$: The scaling function is orthogonal if and only if for any problematic conjugate cycle, at least one point in that cycle is not a zero of the mask function. This implies that for sufficiently large $\Delta$, orthogonality can be determined by checking a finite set of points.
  • If $\Delta < \Delta_0$: There are infinitely many problematic conjugate cycles, and thus orthogonality cannot be determined by checking only a finite set of points in the interval $[-\Delta, \Delta]$. The value $\Delta_0$ is explicitly defined in terms of the limit of the $\mu(D_k)$ sequence, approximately 0.0875.

Limitations and Future Work

The paper focuses on a specific class of mask functions that are trigonometric polynomials and satisfy the Smith-Barnwell condition. While the results are significant for this class, extensions to more general types of mask functions might be a direction for future work. The authors also acknowledge that their work builds upon and refines existing results, suggesting a continuous evolution of understanding in this area. The paper's conclusion implies that for small $\Delta$, the problem of determining orthogonality becomes intractable by simply checking for zeros in a finite set.

Summary updated

BibTeX
@inproceedings{zhou2000zeros,
  author = {Xingwei Zhou and Weifeng Su},
  title = {{Zeros of a Mask Function and the Orthogonality of the Related Scaling Function}},
  booktitle = {Applied and Computational Harmonic Analysis},
  year = {2000},
  doi = {10.1006/acha.2000.0309},
}

Loading related content...

Figure