Finsler-Laplace-Beltrami Operators with Application to Shape Analysis

Simon Weber,Thomas Dagès,Maolin Gao,Daniel Cremers

2024-04-05

Abstract:The Laplace-Beltrami operator (LBO) emerges from studying manifolds equipped with a Riemannian metric. It is often called the Swiss army knife of geometry processing as it allows to capture intrinsic shape information and gives rise to heat diffusion, geodesic distances, and a multitude of shape descriptors. It also plays a central role in geometric deep learning. In this work, we explore Finsler manifolds as a generalization of Riemannian manifolds. We revisit the Finsler heat equation and derive a Finsler heat kernel and a Finsler-Laplace-Beltrami Operator (FLBO): a novel theoretically justified anisotropic Laplace-Beltrami operator (ALBO). In experimental evaluations we demonstrate that the proposed FLBO is a valuable alternative to the traditional Riemannian-based LBO and ALBOs for spatial filtering and shape correspondence estimation. We hope that the proposed Finsler heat kernel and the FLBO will inspire further exploration of Finsler geometry in the computer vision community.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

### Problems the paper attempts to solve

This paper aims to explore the application of Finsler manifolds in shape analysis and proposes a new Finsler - Laplace - Beltrami operator (FLBO). Specifically, the paper addresses the following problems:

1. **Limitations of traditional Riemannian manifolds**:
- Riemannian manifolds are widely used in many computer vision and machine learning tasks, but they assume that the metric is quadratic, which limits their flexibility.
- Finsler manifolds introduce anisotropic anisotropy by removing the quadratic assumption and allowing the metric to depend not only on the position but also on the tangential direction.

2. **Shape correspondence problem**:
- Shape correspondence is a fundamental task in computer vision, which involves finding corresponding points between two shapes.
- The traditional Riemannian Laplace - Beltrami operator (LBO) and its anisotropic version (ALBO) have limitations when dealing with shape correspondence.
- The paper proposes a new Finsler - Laplace - Beltrami operator (FLBO) to improve the accuracy of the shape correspondence task.

3. **Generalization of the heat diffusion equation**:
- The paper re - examines the heat diffusion equation on Finsler manifolds and derives the Finsler heat kernel.
- These theoretical results provide a basis for further exploring the application of Finsler geometry in computer vision.

### Main contributions

1. **Theoretical exploration**:
- Provides a concise and self - contained introduction to Finsler manifold theory, making it easy to be understood by the computer vision and machine learning communities.
- Re - examines the heat diffusion equation on Finsler manifolds and derives a tractable equation.

2. **Proposal of a new operator**:
- Generalizes the traditional (anisotropic) Laplace - Beltrami operator and proposes the Finsler - Laplace - Beltrami operator (FLBO).
- By exploiting the relationship between the diffusion coefficient and the Riemannian metric, proposes an easy - to - implement FLBO discretization method.

3. **Experimental verification**:
- Conducts extensive evaluations of the proposed FLBO on the shape correspondence task, demonstrating its superior performance on full - shape and partial - shape datasets.
- Provides open - source code to promote further research.

### Conclusion

This paper extends the traditional Riemannian manifold theory by introducing the Finsler - Laplace - Beltrami operator (FLBO), providing new tools for shape analysis and geometric deep learning. Experimental results show that FLBO has higher accuracy in the shape correspondence task, especially when dealing with partial - shape data.