Purchase this article with an account.
Barton L. Anderson, Juno Kim, Phillip Marlow; At what level of representation is surface gloss computed?. Journal of Vision 2011;11(11):396. doi: https://doi.org/10.1167/11.11.396.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
We recently argued that the perception of gloss depends on the congruence between computations of 3D shape and the locations of highlights, and showed that displacements of highlights in the image plane destroyed the perception of gloss. Here, a series of experiments were conducted to further assess this congruence hypothesis by manipulating highlight locations in 3D. We rendered bumpy spherical surfaces using natural illumination fields. The specular and diffuse reflectance components were decomposed and stored as separate 3D maps. In Experiment 1, the specular and diffuse shading maps were rotated relative to each other in 3D by varying degrees and then additively combined, introducing “phase shifts” between the diffuse and specular reflectance maps. In Experiment 2, we introduced highlight incongruences by rotating the illumination field of the diffuse reflectance component (only) by varying degrees. In Experiment 3, we combined these two sources of incongruence, introducing both “phase shifts” and illumination field misalignments between the diffuse and specular reflectance components. Observers were shown pairs of stimuli and reported which image appeared glossier, for both static and moving displays. The results of these experiments demonstrated that the percept of gloss decreased monotonically as a function of highlight misalignment. In Experiment 1, the decline in perceived gloss with phase shift misalignment dropped off steeply for small misalignments, and asymptoted quickly, particularly for the moving images. The decline in perceived gloss with the illumination field rotation generated a nearly linear decline over the full 180 degrees of rotation for both static and moving images, and dominated the effects of phase offset when the two sources of incongruence were combined in Experiment 3. The results of these experiments provide strong evidence that the computations responsible for the perception of surface gloss occur in a 3D representational space, which cannot be understood with 2D image statistics.
This PDF is available to Subscribers Only