A psychophysical evaluation of techniques for Mooney image generation

Area of Science:

• Cognitive Psychology
• Computer Vision
• Image Processing

Background:

• Mooney images dissociate image content from understanding, aiding visual perception research.
• Manual generation of Mooney images is subjective; automatic methods are underutilized.
• Limited research exists on perceptual effects of smoothing and thresholding in Mooney image creation.

Purpose of the Study:

• Investigate the impact of thresholding techniques and smoothing on Mooney image interpretability.
• Explore automatic generation methods for Mooney images.
• Assess human participant performance in identifying concepts in generated Mooney images.

Main Methods:

• Generated Mooney images using four global thresholding techniques.
• Parametrically varied smoothing levels in a second experiment.
• Collected participant data on concept identification and subjective interpretability ratings.

Main Results:

• Identification performance and subjective ratings were similar across different global thresholding techniques.
• The specific global thresholding method had minimal impact on Mooney image interpretability.
• The degree of smoothing significantly influenced interpretability, requiring careful adjustment based on image noise and desired outcome.

Conclusions:

• Global thresholding is not critical for Mooney image interpretability.
• Smoothing level is a crucial parameter requiring optimization for automatic Mooney image generation.
• Future research should focus on local thresholding techniques for enhanced automatic generation.