Abstract
We present evidence that the perceived contrast of complex, real-world images is biased towards their high spatial frequency components, and that neither the contrast sensitivity function nor contrast constancy are useful descriptions of broadband contrast perception. We adapted the classification image paradigm to measure decision weighting functions for luminance contrast as a function of spatial frequency in systematically modified broadband natural images. The experiment yields bandpass weighting functions which can be explained by combining an existing model of perceived contrast (Cannon & Fullenkamp 1991) with suppression disproportionately exerted on low spatial frequency responses. We were also able to recover local luminance classification functions showing that negative-polarity ('dark') contrasts are most important to human judgments of overall image contrast, consistent with recent evidence that dark image regions are encoded preferentially during early contrast encoding (e.g. Komban, Alonso, & Zaidi 2011). Our implementation of suppression is simplified but consistent with broader aspects of contrast psychophysics — we predict that studies of contrast masking or adaptation using broadband images should show that suppression is progressively stronger towards low spatial frequencies, keeping perceptual responses just above mechanism thresholds and ensuring that higher-frequency components are perceptually emphasized.
Meeting abstract presented at OSA Fall Vision 2012