Natural image statistics mediate brightness 'filling in'

Area of Science:

• Visual perception
• Computational neuroscience
• Image processing

Background:

• The human visual system perceives surface reflectance despite varying illumination.
• The Craik-Cornsweet-O'Brien (CCOB) effect, an illusion where equiluminant grey regions appear different due to luminance transitions, is often attributed to low-level 'filling-in' mechanisms.

Purpose of the Study:

• To investigate the underlying mechanism of the Craik-Cornsweet-O'Brien (CCOB) effect.
• To challenge the prevailing view of spatial neural signal propagation for CCOB-induced lightness illusions.
• To propose and validate a model based on spatial frequency (SF) structure amplification.

Main Methods:

• Development of a computational model using natural scene statistics to reconstruct images based on SF channel responses.
• Psychophysical testing involving classification images (CIs) to analyze discrimination of CCOB stimuli contrast polarity under noise.
• Analysis of observer reliance on specific stimulus regions in relation to perceptual completion.

Main Results:

• Evidence suggests the CCOB effect mechanism amplifies low spatial frequency (SF) image structure, rather than relying on spatial neural signal propagation.
• Classification images derived from psychophysical experiments resemble 'filled-in' stimuli, indicating reliance on non-informative regions.
• The 'filling-in' process in CCOB illusions is confirmed to be dependent on the presence of suitable low SF structure.

Conclusions:

• The Craik-Cornsweet-O'Brien (CCOB) effect in lightness perception is explained by the amplification of low spatial frequency image components.
• This amplification mechanism, informed by natural scene statistics, provides a more accurate account of the CCOB illusion than spatial propagation models.
• Understanding the role of low SF structure is crucial for comprehending visual filling-in phenomena.