Purchase this article with an account.
Yoana Dimitrova, Peter McOwan, Alan Johnston; Lightness estimation errors in a 3D context. Journal of Vision 2010;10(7):453. doi: https://doi.org/10.1167/10.7.453.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The problem of recovering reflectance from a single image is inherently under-constrained. Therefore, the visual system must use heuristics or biases to recover reflectance, whilst discounting geometry and illumination. In a previous study, we investigated lightness perception in a 3D context (Dimitrova YD, McOwan P, Johnston A, 2009, Perception 38: ECVP Abstract supplement, p. 31). Participants systematically overestimated reflectance for vertically eccentric illuminant angles and underestimated for illuminant angles close to the horizontal plane. These errors were robust to additional cues to light source direction and depth as well as to the removal of local features. Errors rose significantly with the increase in the ratio of directional to ambient light. Modelling of the data indicated three possible causes of the reflectance misestimation: a light direction bias, a bias in the proportion of ambient to directional light or a simple brightness averaging over the image. To investigate perceived illuminant direction, participants were asked to adjust the illumination direction for a sphere until it matched the lighting direction for a dodecahedron rendered using a range of illuminant elevations. The pattern of light direction adjustments was consistent with a bias of perceived illuminant elevation that is both shifted away from vertically eccentric angles and away from the horizontal plane. These adjustments were similar the results of modelling reflectance errors with a light direction bias. This finding supports the view that the systematic errors in reflectance settings are at least in part caused by a bias in the assumed direction of illumination.
This PDF is available to Subscribers Only