Image statistics and the perception of surface qualities

Area of Science:

• Visual perception
• Computer vision
• Material science

Background:

• Real-world surfaces possess complex properties like gloss and mesostructure.
• Understanding surface appearance involves disentangling illumination, reflectance, and geometry from images.
• Existing research often focuses on simplified, textureless matte surfaces.

Purpose of the Study:

• Investigate simple image statistics that correlate with surface material properties.
• Determine if human observers utilize these statistics for surface judgments.
• Explore the neural mechanisms underlying surface appearance perception.

Main Methods:

• Analysis of luminance histogram skewness and sub-band filter outputs.
• Correlation analysis between image statistics and surface properties (gloss, albedo).
• Psychophysical experiments including visual adaptation to skewed stimuli.

Main Results:

• Luminance histogram skewness correlates positively with surface gloss and inversely with albedo.
• Human observers demonstrate a bias towards judging surfaces with higher skewness as darker and glossier.
• Visual adaptation to skewed patterns alters perceived lightness and glossiness.

Conclusions:

• Skewness of image statistics serves as a diagnostic cue for surface gloss and albedo.
• Human visual system likely employs skewness-based mechanisms for surface property estimation.
• These findings suggest neural sensitivity to statistical properties of visual input for perception.