Ensemble Coding in Color and Blur Perception
StatisticsView Usage Statistics
At any moment sensory systems are confronted with vast amounts of information, and therefore must represent information as efficiently as possible. In recent years, many studies have proposed ensemble coding as one strategy for forming a compact perceptual code. In a typical visual scene, both color and blur vary widely, and the average has important implications for representing visual norms (e.g. gray or focused) and visual function (e.g. to estimate the illuminant for color or the state of accommodation for blur). Yet whether and how the visual system uses an ensemble code for these features has received little attention. In this work, the properties of ensemble coding for the attributes of color and spatial blur were examined using behavioral and electroencephalogram (EEG) techniques. In a series of experiments, the sensitivity to this average in color and blur was studied using standard member identification and mean discrimination techniques used in summary statistics studies. The visual system has been reported to be biased towards aspects of a scene that are more meaningful to perception. In blur perception, a sharp/focused image is more relevant to vision than blurred images. In another set of experiments, adaptation to blur was studied to examine if the aftereffect is driven by the average blur level or biased towards blurred or sharper parts of the image. Finally, studies on color ensemble coding have focused on average color estimation in a visual scene. In the last set of experiments, aspects of color ensemble coding beyond the average were examined by probing the number of color levels that observers can distinguish in ensembles using both behavioral reaction times and an EEG oddball paradigm. In spite of important perceptual similarities, results from these experiments show that while color ensemble coding happens within opponent hues (e.g. separately for red and green), blur ensembles coding can happen even if an ensemble spans the whole blur spectrum from very sharp to very blurred. This work also shows that ensemble perception is unaffected by specific properties of the ensemble, or by whether or not an ensemble mimics natural variations of color and blur.