Continuous Shearlet Frames and Resolution of the Wavefront Set

Philipp Grohs

Abstract

In the field of multidimensional signal processing, directional multiscale transformations such as curvelets and shearlets have emerged as powerful alternatives to traditional wavelets. While wavelets excel at identifying point-like singularities, they struggle to efficiently represent features along edges. A landmark result by Kutyniok and Labate (2009) established that shearlets can resolve the Wavefront Set—identifying both the location and orientation of singularities—provided the shearlet function $\psi$ possesses frequency support in a specific compact conical wedge. However, this requirement is restrictive for practical applications. This paper demonstrates that the Wavefront Set can be resolved under significantly weaker assumptions, specifically by requiring $\psi$ to possess sufficiently many anisotropic vanishing moments. By adapting the Radon transform to the shearlet structure, we provide a flexible framework for constructing continuous frames in $L^2(\mathbb{R}^2)$ that are suitable for numerical implementation.

Introduction

Traditional wavelet analysis has revolutionized signal processing by providing optimal representations for functions with point-wise singularities. However, in two or more dimensions, interesting data often contains "directional" features, such as edges in an image or surfaces in a volumetric scan. Because wavelets are isotropic—meaning they scale uniformly in all directions—they are inefficient at capturing these anisotropic features, requiring a large number of coefficients to represent even a simple smooth curve.

Shearlets were developed to address this limitation. They utilize a parabolic scaling law (where width $\sim$ length$^2$) and a shearing operation to achieve high directional sensitivity. A fundamental theoretical benchmark for any such transform is its ability to resolve the Wavefront Set of a distribution. The Wavefront Set is a mathematical tool that describes not just where a function is non-smooth, but also the specific directions in which the lack of smoothness occurs.

The Problem: Frequency Constraints vs. Practicality

Previous proofs regarding the shearlet transform's ability to resolve the Wavefront Set relied on very specific frequency-domain constraints. Specifically, the shearlet $\psi$ was required to have support in a "conical wedge." While mathematically elegant, this constraint is problematic for practitioners: functions with such specific frequency support cannot be compactly supported in the spatial domain. In real-world numerical applications, compact support is highly desirable for computational efficiency and localization.

Philipp Grohs addresses this gap by asking: What is the bare minimum requirement for a shearlet to resolve the Wavefront Set? The goal is to move away from rigid frequency-domain shapes toward more general properties that allow for a wider variety of shearlet functions, including those that are compactly supported.

Methodology: Anisotropic Vanishing Moments and the Radon Transform

The core innovation of this work is the shift from frequency-domain geometry to the concept of anisotropic vanishing moments. A function is said to have $n$ vanishing moments if its integrals against polynomials up to degree $n-1$ are zero. Grohs defines a specific version of this property for the $x_1$-direction, which proves to be the critical requirement for directional sensitivity.

To prove these results, the paper develops a new approach based on an adaptation of the Radon transform. By utilizing the Projection Slice Theorem, the author links the 2D shearlet transform to a univariate wavelet transform of the Radon transform. This connection is significant because it reduces the complex 2D problem of singularity detection into a series of 1D problems, which are much easier to analyze using established wavelet theory.

Furthermore, the paper employs the theory of continuous frames. This ensures that the generalized shearlets—even with weakened assumptions—can still stably represent and reconstruct any square-integrable function ($L^2(\mathbb{R}^2)$).

Key Results

1. Resolution of the Wavefront Set

The paper proves (Theorems 1.8 and 6.4) that for any Schwartz function $\psi$ with infinitely many vanishing moments in the $x_1$-direction, the shearlet transform perfectly characterizes the Wavefront Set of a tempered distribution. This confirms that the decay rate of shearlet coefficients acts as a precise "thermometer" for detecting singularities and their orientations.

2. Flexibility and Compact Support

The results hold even if the function only has finitely many vanishing moments, although the resolution is then limited to a specific degree of regularity (referred to as $N$-regularity). This "freedom of choice" means that researchers can use tensor-product wavelets or other practical, compactly supported functions as shearlets without losing the ability to analyze microlocal properties.

3. Tight Frame Construction

The author demonstrates how to construct "tight frames" using these generalized shearlets. A tight frame allows for easy reconstruction of a signal from its coefficients, similar to an orthonormal basis, but with the added flexibility of redundancy. This construction is shown to be valid even when restricted to a frequency cone.

Discussion and Future Work

While this work significantly broadens the theoretical foundation of shearlets, several challenges remain. The current paper focuses on the continuous shearlet transform. Transitioning these results to a discrete setting—specifically creating discrete, non-bandlimited tight frames—remains a difficult task.

Additionally, the author suggests that the definition of the $N$-Wavefront Set used here might be further refined to better correspond with specific microlocal function spaces, such as Besov or Sobolev spaces. Finally, there is the open question of whether these flexible assumptions can be applied to rotation-based systems, such as curvelets, which share similar goals but use different geometric transformations.

By relaxing the requirements for shearlet construction, this research bridges the gap between high-level microlocal analysis and practical signal processing, providing a rigorous justification for the use of shearlets in a much wider array of applications.

Summary updated

Abstract

In recent years directional multiscale transformations like the curvelet- or shearlet transformation have gained considerable attention. The reason for this is that these transforms are - unlike more traditional transforms like wavelets - able to efficiently handle data with features along edges. The main result in [G. Kutyniok, D. Labate. Resolution of the Wavefront Set using continuous Shearlets, Trans. AMS 361 (2009), 2719-2754] confirming this property for shearlets is due to Kutyniok and Labate where it is shown that for very special functions $ψ$ with frequency support in a compact conical wegde the decay rate of the shearlet coefficients of a tempered distribution $f$ with respect to the shearlet $ψ$ can resolve the Wavefront Set of $f$. We demonstrate that the same result can be verified under much weaker assumptions on $ψ$, namely to possess sufficiently many anisotropic vanishing moments. We also show how to build frames for $L^2(\mathbb{R}^2)$ from any such function. To prove our statements we develop a new approach based on an adaption of the Radon transform to the shearlet structure.
BibTeX
@inproceedings{grohs2010continuous,
  author = {Philipp Grohs},
  title = {{Continuous Shearlet Frames and Resolution of the Wavefront Set}},
  booktitle = {Monatshefte},
  year = {2010},
  doi = {10.1007/s00605-010-0264-2},
}
Figure