Jiayi Kong,Chen Zong,Jun Luo,Shiqing Xin,Fei Hou,Hanqing Jiang,Chen Qian,Ying He
Abstract:The medial axis, a lower-dimensional shape descriptor, plays an important role in the field of digital geometry processing. Despite its importance, robust computation of the medial axis transform from diverse inputs, especially point clouds with defects, remains a significant challenge. In this paper, we tackle the challenge by proposing a new implicit method that diverges from mainstream explicit medial axis computation techniques. Our key technical insight is the difference between the signed distance field (SDF) and the medial field (MF) of a solid shape is the unsigned distance field (UDF) of the shape's medial axis. This allows for formulating medial axis computation as an implicit reconstruction problem. Utilizing a modified double covering method, we extract the medial axis as the zero level-set of the UDF. Extensive experiments show that our method has enhanced accuracy and robustness in learning compact medial axis transform from thorny meshes and point clouds compared to existing methods.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: **Robustly compute the compact Medial Axis Transform (MAT) from defective input data (such as point clouds with noise or partial missing)**. Specifically, existing medial axis computing methods face challenges when dealing with shapes of complex topological structures and irregular surfaces, especially when the input data quality is not high (such as when point clouds are defective). This paper proposes a new implicit method - Quasi - Medial Distance Field (Q - MDF) to overcome these challenges.
### Problem Background
The medial axis is a classic geometric structure that can capture the local symmetry and global topological features of a shape. When combined with radius information to form the Medial Axis Transform (MAT), it can be used to reconstruct the original shape and serve as an important low - dimensional shape descriptor. However, the medial axis is not a stable structure. Small perturbations on the boundary may introduce additional branches, increasing its complexity and thus affecting the performance of downstream applications.
In addition, the input data are not always perfect closed meshes or densely sampled point clouds. Common point clouds may have non - uniform distribution, noise, or partially missing areas, which increases the difficulty of inferring a reasonable original shape from these data. Therefore, although the medial axis has a strict definition, robustly computing a compact medial axis remains a challenge.
### Limitations of Existing Methods
- **Traditional methods**: Most traditional methods rely on the discretization of the shape or the medial axis (such as voxelization or obtaining medial axis points). Although these methods are simple and efficient, their accuracy is limited by the resolution and they are difficult to handle complex input data.
- **Learning - based methods**: Although they reduce the requirement for high - quality input, they face challenges in generalization ability and maintaining centrality and topological correctness.
### Solutions Proposed in the Paper
This paper proposes the Q - MDF method, which solves the above problems through the following steps:
1. **Implicit representation**: Utilize the relationship between SDF (Signed Distance Field) and MF (Medial Field) to construct a new differentiable continuous representation - Q - MDF, which is similar to the unsigned distance field (UDF) of the medial axis but easier to compute.
2. **Medial axis extraction**: Use an improved double - covering method to extract the medial axis from Q - MDF as a zero - level set to obtain a compact Medial Axis Transform.
3. **Enhancing sharp features**: For shapes with sharp features (such as CAD models), further enhance the representation of these features, making the extracted medial axis more accurate and refined.
### Main Contributions
1. Propose a new approximation method for the Medial Axis Transform, which can robustly generate a compact Medial Axis Transform from complex input data.
2. Based on the relationship between SDF and MF, propose a differential representation of the medial axis.
3. Utilize the double - covering concept to develop an algorithm for accurately extracting the medial axis mesh from the differential medial axis representation.
4. For the sharp features of the original shape, further enhance the representation of the corresponding areas, making the extracted medial axis more accurate and refined.
Through these innovations, this paper provides a more robust and more general medial axis computing method, which is suitable for various complex scenarios.