A remark on the anisotropic outer Minkowski content

Antonin Chambolle, Stefano Lisini, Luca Lussardi

Abstract. We study an anisotropic version of the outer Minkowski content of a closed set in $\mathbb{R}^n$. In particular, we show that it exists on the same class of sets for which the classical outer Minkowski content coincides with the Hausdorff measure, and we give its explicit form.

Introduction

In geometric measure theory, the "size" of a boundary is typically quantified using the Hausdorff measure. However, the Minkowski content provides a powerful alternative by relating the surface area of a set to the volume of a thin "skin" or neighborhood surrounding it. Classically, this is an isotropic measurement: it treats all directions equally by using standard Euclidean balls to define the neighborhood.

This paper extends the concept of outer Minkowski content to an anisotropic setting. By replacing the standard ball with a general convex body $C$, the authors demonstrate that the resulting measure converges to the expected anisotropic perimeter for a wide class of rectifiable sets. This extension is particularly relevant for modeling physical phenomena with directional preferences, such as crystal growth, flame propagation, and advanced image processing.

Defining the Anisotropic "Skin"

The core challenge addressed in this research is twofold: rigorously defining an "outer" Minkowski content using an arbitrary convex shape $C$ and determining the conditions under which this definition converges to the theoretical anisotropic perimeter—a weighted surface area where the weight depends on the direction of the surface normal.

To handle this anisotropy, the authors utilize the support function $h_C$ of a convex body $C$. The support function describes the "reach" of the shape in any given direction, allowing for the weighting of the surface normal $\nu_E$ accordingly. By focusing specifically on the outer Minkowski content—measuring only the volume added outside the set—the authors simplify the relationship to the perimeter of sets with finite perimeter, bypassing complexities associated with the two-sided classical version.

Methodology

The study employs several sophisticated tools from the calculus of variations and geometric measure theory:

  • $Γ$-convergence: Rather than focusing solely on pointwise limits, the authors analyze the convergence of the Minkowski content as a functional. This provides a robust framework for understanding how these measures behave when applied to sequences of sets.
  • Blow-up Analysis: To prove the primary convergence theorems, the authors perform a "zoom-in" on points of the reduced boundary. By demonstrating that the boundary appears increasingly flat at infinitesimal scales, they can effectively compare the volume of the anisotropic neighborhood to the area of a flat hyperplane.
  • Co-area Formula: A generalized co-area formula is used to bridge the gap between the content of sets and the content of functions, specifically functions of Bounded Variation ($BV$).

Key Results and Insights

Consistency with Classical Theory

A primary contribution of the paper is Theorem 3.4 (Pointwise Convergence). It establishes that if a set is sufficiently "well-behaved" for the standard isotropic Minkowski content to exist and match the perimeter, it is automatically suitable for the anisotropic version. Specifically, the anisotropic limit is shown to be exactly: $$\int_{\partial^* E} h_C(\nu_E) d\mathcal{H}^{n-1}$$

Functional Approximation

The authors prove in Theorem 3.1 that the anisotropic outer Minkowski content $\Gamma$-converges to the anisotropic perimeter $Per_{h_C}(E; \Omega)$. This ensures that minimizers of the approximate functionals converge to the minimizers of the actual perimeter. Furthermore, Corollary 3.2 extends this to functions, showing that a specific functional involving local suprema converges to the anisotropic total variation of a $BV$ function.

This functional approximation has significant implications for numerical methods. In fields like image processing, where anisotropic total variation is a common tool for noise reduction and edge detection, these results provide a theoretical foundation for new approximation algorithms.

Limitations and Future Directions

While the paper establishes a rigorous foundation, several avenues for future research remain:

  1. Smoothness of the Convex Body: The equivalence of certain assertions in Theorem 3.4 may require the convex body $C$ to be $C^{1,1}$ and uniformly convex (elliptic). The behavior of the content when $C$ is non-smooth remains an open question.
  2. Fractal Boundaries: The current results rely on the boundary being rectifiable. The behavior of anisotropic content on sets with fractal boundaries—where classical Minkowski content often fails to exist—is not yet addressed.
  3. Applied Evolutions: The authors suggest these results are motivated by stochastic geometry. Applying these specific anisotropic limits to models of anisotropic mean curvature flow could yield new insights into physical growth processes.

Conclusion

By generalizing the outer Minkowski content to include directional weights, this work bridges the gap between isotropic geometric measures and the anisotropic perimeters required for realistic physical modeling. The proof that the anisotropic version exists under the same conditions as the classical one provides a powerful, unified tool for geometric analysis.

Summary updated

Abstract

We study an anisotropic version of the outer Minkowski content of a closed set in <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"> <m:msup> <m:mrow> <m:mi>ℝ</m:mi> </m:mrow> <m:mi>n</m:mi> </m:msup> </m:math> ${\mathbb {R}}^n$ . In particular, we show that it exists on the same class of sets for which the classical outer Minkowski content coincides with the Hausdorff measure, and we give its explicit form.
BibTeX
@inproceedings{chambolle2013remark,
  author = {Antonin Chambolle and Stefano Lisini and Luca Lussardi},
  title = {{A remark on the anisotropic outer Minkowski content}},
  booktitle = {Advances},
  year = {2013},
  doi = {10.1515/acv-2013-0103},
}
Figure