Spectral Descriptors for 3D Deformable Shape Matching: A Comparative Survey.

Area of Science:

• Computer Vision
• Geometric Processing
• Shape Analysis

Background:

• 3D spectral descriptors are crucial for deformable shape matching.
• Key descriptor qualities include descriptive capacity, storage efficiency, and robustness.
• Current literature lacks comprehensive comparisons of existing spectral descriptors.

Purpose of the Study:

• To comprehensively evaluate nine state-of-the-art 3D spectral descriptors.
• To assess descriptor performance across various datasets and perturbations.
• To provide guidance for selecting and developing spectral descriptors.

Main Methods:

• Evaluation of nine spectral descriptors on ten deformable shape datasets.
• Testing robustness against mesh discretization, geometric noise, scale, non-isometry, partiality, and topological noise.
• Assessment based on distinctiveness, robustness, compactness, and computational efficiency.

Main Results:

• Detailed performance analysis of nine spectral descriptors under diverse conditions.
• Identification of strengths and weaknesses of each descriptor.
• Quantification of descriptor performance across multiple evaluation metrics.

Conclusions:

• Summary of overall performance and key findings.
• Guidance for researchers on choosing appropriate spectral features for specific applications.
• Insights for the development of novel and improved 3D spectral descriptors.