Sketch Matching on Topology Product Graph

Area of Science:

• Computer Science
• Human-Computer Interaction
• Artificial Intelligence

Background:

• Sketch matching is crucial for sketch-based interfaces but is hindered by variations in hand-drawn shapes.
• Existing methods often rely on limited topology exploration and suboptimal graph similarity metrics.
• The challenge lies in accurately matching sketches with significant irregularities and variations.

Purpose of the Study:

• To develop a robust sketch matching method addressing limitations of current approaches.
• To introduce a comprehensive definition of topology relations for richer sketch representation.
• To propose a novel graph matching technique and similarity metric for improved accuracy and partial matching capabilities.

Main Methods:

• Developed a comprehensive definition of topology relations to create informative graph representations of sketches.
• Introduced the topology product graph for retaining full correspondence during graph matching.
• Derived an intuitive and easily computable sketch similarity metric based on the topology product graph.

Main Results:

• The proposed method generates a rich graph representation capturing detailed topology relations.
• The topology product graph facilitates accurate matching by preserving full correspondence.
• The new similarity metric enables efficient computation and supports partial sketch matching.
• Experimental results demonstrate superior performance compared to state-of-the-art methods on a challenging dataset.

Conclusions:

• The novel sketch matching method significantly advances the field by offering improved accuracy and handling of variations.
• The comprehensive topology relations and topology product graph provide a more robust framework for sketch recognition.
• The method's ability to support partial matching addresses a critical practical need in sketch-based interfaces.