A Compact Shape Descriptor for Triangular Surface Meshes.

Area of Science:

• Computer Vision
• Geometric Modeling
• Computational Geometry

Background:

• 3D shape descriptors are crucial for object recognition and database retrieval.
• Existing methods like Shape-DNA offer isometric-invariance but can be computationally intensive due to numerous eigenvalues.
• Large-scale applications necessitate more compact and efficient shape representation techniques.

Purpose of the Study:

• To develop a novel, compact shape descriptor for triangular surface meshes.
• To reduce the computational and storage requirements of existing shape description methods.
• To enhance the efficiency of 3D shape analysis, comparison, and classification.

Main Methods:

• The standard Shape-DNA descriptor is computed for a given mesh.
• The computed Shape-DNA undergoes surface area-based normalization and line subtraction.
• The compact Shape-DNA (cShape-DNA) is derived from low-frequency components of the discrete Fourier transform of the processed descriptor.

Main Results:

• The proposed cShape-DNA method effectively describes triangular surface mesh shapes.
• Experiments demonstrate the effectiveness and efficiency of cShape-DNA in 3D shape analysis.
• The method shows particular promise for shape comparison and classification tasks.

Conclusions:

• cShape-DNA offers a more compact and computationally efficient alternative to standard Shape-DNA.
• The descriptor is well-suited for large-scale 3D shape database applications.
• The method provides a valuable tool for advanced 3D shape analysis and recognition.