Good continuation in dot patterns: A quantitative approach based on local symmetry and non-accidentalness

Area of Science:

• Computer Vision
• Perceptual Psychology
• Computational Neuroscience

Background:

• Grouping principles, like good continuation, are fundamental to visual perception.
• Previous models often struggle with robustness and scale invariance in pattern detection.
• Local symmetries and non-accidentalness offer promising cues for perceptual organization.

Purpose of the Study:

• To introduce a novel computational model for grouping dot patterns based on the good continuation law.
• To develop a quantitative measure of non-accidentalness for predicting perceptual relevance.
• To create a robust, unsupervised, and scale-invariant algorithm for detecting good continuation in dot patterns.

Main Methods:

• Utilizing local symmetries and the principle of non-accidentalness to identify perceptually relevant configurations.
• Proposing a quantitative measure of non-accidentalness.
• Deriving a robust, unsupervised, and scale-invariant algorithm for good continuation detection.

Main Results:

• The proposed quantitative measure of non-accidentalness shows good correlation with the visibility of dot curves.
• The developed algorithm successfully detects good continuation in various datasets.
• The method's performance is validated against classic psychophysical study data.

Conclusions:

• The novel approach effectively models the good continuation law for dot pattern grouping.
• The algorithm provides a robust and scalable solution for perceptual organization tasks.
• The findings contribute to understanding visual grouping mechanisms and have practical applications in computer vision.