A computational model of the short-cut rule for 2D shape decomposition

Area of Science:

• Computer Vision
• Cognitive Science
• Computational Geometry

Background:

• Human visual perception favors partitioning objects using minimal path lengths.
• Existing shape decomposition methods lack intuitive human-like part segmentation.

Purpose of the Study:

• To develop and implement a computational model of the short-cut rule for 2D shape decomposition.
• To validate the model's ability to generate human-intuitive part segmentations.

Main Methods:

• A computational model based on the short-cut rule was developed.
• Cut hypotheses were generated at negative curvature minima on object silhouettes.
• Hypotheses were pruned using local property analysis and ranked by length.

Main Results:

• The proposed method effectively decomposes 2D shapes into intuitive parts.
• Results demonstrate superior alignment with human perceptual judgments compared to existing methods.
• The short-cut rule provides a robust framework for part-based shape representation.

Conclusions:

• The short-cut rule offers a biologically plausible and effective approach to computational shape decomposition.
• This method advances the field of computer vision by integrating cognitive principles.
• Future work can explore extensions to 3D shape analysis and real-world object recognition.