Abstract
Human observers perceive the lightness of surfaces accurately despite substantial variations in illumination and viewing conditions. It is known that the lightness of an image region is influenced by its surroundings, but the exact mechanism for this contextual modulation is not yet understood. We recently proposed a contrast-based lightness model (Zeiner & Maertens, 2014) to predict the perceived surface lightness from retinal image luminance. However, this type of model does not account for so-called assimilation effects, i.e. substantial lightness differences that have been reported in the absence of differences in either luminance or contrast. Here, we assessed the magnitude of contrast and assimilation effects in two types of stimuli that differed in photometric and geometric complexity. The more complex stimulus was a two-dimensional image of a checkerboard that was shown as a perspective projection of a three-dimensional object. It was composed of a number of different luminance values and contained a number of cues to depth. The less complex stimulus was a simultaneous lightness contrast type of display that was composed of image regions that emulated the luminances in the checkerboard but lacked its depth structure. Contrary to previous studies we observed strong assimilation effects. In the checkerboard assimilation and contrast effects were of similar magnitude. In addition, assimilation effects seemed stronger for incremental targets on incremental checks. This suggests a split between increments and decrements which is supported by neurophysiological data pointing to a fundamental separation between neurons driven by darks (OFF neurons) and neurons driven by lights (ON neurons) in primary visual cortex (e.g. Yeh et al., 2009; Kremkow et al., 2014). A comprehensive theory of surface lightness must explain why assimilation is so strong on the checkerboard as well as the large asymmetry between black and white.
Meeting abstract presented at VSS 2015