Purchase this article with an account.
Anya C. Hurlbert, Yazhu Ling; Color constancy of chromatically textured surfaces. Journal of Vision 2006;6(6):247. doi: 10.1167/6.6.247.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Computational models of color constancy demonstrate (implicitly or explicitly) that the estimation of the illumination spectral power distribution necessarily improves as the number of distinct surface reflectance samples increases. Although the underlying assumption of such models is that each distinct surface is uniform in reflectance, we observe that a single surface with intrinsic chromatic texture may provide a large number of reflectance samples on its own. But chromatic texture within a surface also blocks simultaneous chromatic contrast between the surface and its background, most powerfully when the background is uniform (Hurlbert & Wolf, Prog. Brain Res., 2003). Thus, the contribution of local between-surface contrast to color constancy for chromatically textured surfaces is weak at best, unlike the empirical results for artificial, homogeneous surfaces (e.g. Kraft & Brainard, Proc. Nat. Sci. Acad., 1999). The contribution of within-surface chromatic contrast, though, may be strong, depending on the spatial scale and luminance statistics of the texture. Here we record and analyze the chromatic texture of representative natural objects and model its effect on color constancy, by estimating limits on the contributions from (1) within-surface chromatic contrast (2) between-surface chromatic contrast (for homogeneous backgrounds) and (3) illumination estimation based on within-surface reflectance sampling. We predict that surface color constancy may be better for objects with natural chromatic texture than for the homogeneous surfaces typically used in laboratory measurements of constancy.
This PDF is available to Subscribers Only