Visual discrimination of textures with identical third-order statistics

Area of Science:

• Computer Vision
• Image Processing
• Statistical Modeling

Background:

• Texture analysis often relies on statistical properties like power spectra (second-order statistics).
• It is commonly assumed that texture granularity is determined by second-order statistics.
• Higher-order statistics are typically used for more complex texture discrimination.

Purpose of the Study:

• To identify and characterize a new class of two-dimensional random textures.
• To investigate the perceptual discrimination of textures with identical third-order statistics.
• To challenge the conventional understanding of texture granularity determination.

Main Methods:

• Generation of novel two-dimensional random textures.
• Analysis of third-order statistics for texture characterization.
• Perceptual experiments to assess texture discrimination based on local granularity.

Main Results:

• A new class of two-dimensional random textures with identical third-order statistics was identified.
• These textures were found to be effortlessly discriminated based on local granularity differences.
• Visible texture granularity was shown to be independent of power spectra and third-order statistics.

Conclusions:

• Texture granularity is not solely determined by power spectra (second-order statistics) or third-order statistics.
• Local granularity differences can be a primary factor in texture discrimination, even when higher-order statistics are identical.
• This finding necessitates a re-evaluation of texture analysis models and perception theories.