Photometric determinants of perceived transparency
Manish Singh and Barton L. Anderson
Vision Research, 46, 879-894 (2006)

Abstract

Photometric constraints for the perception of transparency were investigated using stereoscopic textured displays. A contrast discontinuity divided the textured displays into two lateral halves, with one (reference) half fixed. Observers adjusted the luminance range within the other (test) half in order to perform two tasks: (i) indicate the highest luminance range for which the test side is perceived to be transparent, and (ii) indicate the lowest luminance range for which the test side is seen as being in plain view. Settings were obtained for multiple values of test mean luminance, in order to map out the perceptual locus of transition between transparency and non-transparency. The results revealed a systematic violation of Metelli's magnitude constraint in predicting the percept of transparency. Observer settings were approximated instead by a constraint based on perceived contrast (which matched Michelson contrast for the textures used). The results also revealed large asymmetries between darkening and lightening transparency. When the test was darker than the reference, settings were highly consistent across observers and closely followed the Michelson-contrast prediction. When the test was lighter, however, there was greater variability across observers, with two observers exhibiting shifts toward Metelli's magnitude constraint. Moreover, each observer's setting reliability was significantly worse for lightening transparency than darkening transparency. These results suggest that (polarity-preserving) darkening serves as an additional cue to perceptual transparency.