Related Experiment Videos
Textural properties corresponding to visual perception based on the correlation mechanism in the visual system.
Kenji Fujii1, Shinofu Sugi, Yoichi Ando
1Ando Laboratory, Graduate School of Science and Technology, Kobe University, Rokkodai, Nada, Kobe, 657-8501 Japan. fujii@ymec.com
Psychological Research
|September 5, 2003
Summary
Autocorrelation function (ACF) analysis quantifies visual texture perception. This method effectively measures texture contrast, coarseness, and regularity, aligning computational data with human perception for machine vision applications.
Area of Science:
- Computer Vision
- Perception Science
- Image Analysis
Background:
- Computational measurements of visual texture are crucial for machine vision and computer interfaces.
- Understanding how the human visual system extracts texture information is key to developing effective computational models.
- Existing methods may not fully capture the perceptual properties of visual textures.
Purpose of the Study:
- To present texture parameters derived from autocorrelation function (ACF) analysis that correlate with perceptual properties.
- To demonstrate the utility of ACF analysis for quantifying texture contrast, coarseness, and regularity.
- To validate the ACF analysis by comparing its outputs with subjective human judgments of natural textures.
Main Methods:
- Utilized autocorrelation function (ACF) analysis to extract texture features.
- Analyzed the structure of the estimated ACF, including periodical peaks and decay rates.
- Collected subjective scores for perceived texture properties (contrast, coarseness, regularity) from human observers.
- Compared ACF-derived parameters with subjective scores to assess validity.
Main Results:
- ACF analysis successfully provided measures for texture contrast, coarseness, and regularity.
- For textures with harmonic structures, periodical peaks in the ACF corresponded to perceived coarseness and regularity.
- For random textures, the decay rate of the ACF effectively represented coarseness and regularity.
- Calculated ACF factors showed strong correlation with subjective perceptual scores.
Conclusions:
- Autocorrelation function (ACF) analysis is a valid and effective method for quantifying perceptual properties of visual texture.
- The structure of the ACF (periodical peaks or decay rate) provides distinct information about texture characteristics.
- This approach enhances machine vision capabilities by aligning computational texture analysis with human perception.