Abstract
Transparent color filters have the fascinating quality of appearing more like lights than surfaces. Despite our eyes receiving light that is a multiplicative composite of surface, illumination, and filter properties, we can simultaneously perceive surface and overlay layers differing in colors. How neurons separate the information to extract the transparent layer remains unknown. Physical characteristics of transparent filters generate geometrical and color features in retinal images which could provide cues for separating layers. We estimated the relative importance of such cues in a perceptual scale for transparency, using stimuli in which X or T-junctions, consistent or inconsistent motions, and consistent or inconsistent colors, cooperated or competed in forced-preference psychophysics experiments with 5 observers (72 repeats/pair). Bayesian Thurstone scaling on observer’s preference matrices gave correlations between observers ranging from 0.87 to 0.95, allowing us to combine the observers into a perceptual scale for transparency. This scale revealed some new results: Moving X-junctions increased transparency compared to static X-junctions, but moving T-junctions decreased transparency compared to static T-junctions by creating an opaque patch. However, if the motion of a filter uncovered a dynamically changing pattern having the same motion as the background (common fate) but forming T-junctions, the probability of seeing transparency was second highest after moving X-junctions, despite the stimulus being physically improbable. In addition, geometric cues overrode color inconsistency to a great degree. To uncover latent factors that enhance and inhibit transparency, we built a model of transparency perception as a function of relative motions between filter, overlay, and surround, contour continuation, and color consistency, which generated a scale correlating 0.98 with the combined observer’s scale. This model quantified a hierarchy of latent factors that influence the filter being seen as a separate transparent layer, with the highest weight given to common fate between the overlaid background and the surround.