Scale Space Graph Representation and Kernel Matching for Non Rigid and Textured 3D Shape Retrieval

Area of Science:

• Computer Vision
• Geometric Modeling
• Machine Learning

Background:

• 3D object retrieval is crucial for many applications.
• Existing methods often struggle with variations in scale, texture, and geometry.
• A robust and flexible shape representation is needed.

Purpose of the Study:

• To introduce a novel framework for 3D object retrieval.
• To develop a robust and flexible shape representation method.
• To improve the accuracy and efficiency of 3D shape recognition and retrieval.

Main Methods:

• Utilizing tree-based shape representations (TreeSha) derived from the Auto Diffusion Function (ADF) scale-space.
• Employing specialized graph kernels for comparing these tree-based representations.
• Coupling ADF maxima with basins of attraction to encode multi-scale spatial relationships.
• Incorporating texture and geometric information as node and edge features within the graph structure.

Main Results:

• The TreeSha framework demonstrates competitive or superior retrieval scores compared to state-of-the-art methods on benchmark datasets.
• The method proves effective for both textured and non-textured 3D shape retrieval.
• Experimental results provide insights into the efficacy of different shape descriptors and graph kernels.

Conclusions:

• The proposed TreeSha framework offers a powerful and flexible approach to 3D object retrieval.
• The method's ability to handle multi-scale information and incorporate diverse features enhances its applicability.
• This framework advances the field of 3D shape recognition and retrieval.