Abstract
There is considerable evidence about how color and luminance edge cues are processed in isolation. However, it is less clear how they are combined, particularly when they conflict. Here we compare observed data with predictions from simple maximum likelihood estimation (MLE) based models. We find that chromatic cues were weighted surprisingly strongly in edge localization, given their relative reliability in isolation. Stimuli were comprised of achromatic or isoluminant bipartite fields; step-functions from black to white or red to green, presented in isolation or together. The isoluminant stimuli were designed to isolate the red-green (L-M) channel. Conflicting edges had a gap of 3 minutes of arc between the two step functions, but appeared fused. The contrast of the achromatic component was constant across conditions (0.02) and the isoluminant component was presented at three different contrasts (0.1, 0.2, 0.3). We used a method of adjustment, with unconstrained fixation and unlimited duration, to measure the perceived location of the fused edge and the variance of that judgment. Measurements from the achromatic and isoluminant stimuli were used to predict perceived location in the cue conflict condition. As expected, the reliability of the chromatic cue increased as a function of contrast and increasingly influenced perceived location of the fused edge. MLE accurately predicted the pattern of results across contrasts. However, it consistently over-estimated the relative importance of the luminance cue; this could be accounted for by a simple scale factor. Our results show that the weights generated from measurements of each component in isolation are not sufficient to predict edge location in conflicting conditions. The chromatic component requires a higher weight than would be predicted by MLE. At contrasts for which the components cues are equally reliable for localizing the edge, the visual system gives more weight to the position predicted by the chromatic information.
Meeting abstract presented at VSS 2013