Abstract
We have previously shown that the quality of colour constancy in real scenes is influenced by properties of surface material, namely by the degree of mesostructure. Material mesostructure and appearant roughness are strongly influenced by 3D shape, as is reflexion from non-lambertian surfaces. We asked therefore how these form-material interactions influence colour constancy. We tested colour constancy for different materials (paper, fur, plastic), which were presented as (1) 3D objects of arbitrary forms, (2) well known 3D forms, and (3) flat surfaces. All surfaces / objects extended approximately the same visual angle (15 deg), the flat surfaces were presented at an angle of 60 deg. The samples were mounted on black sample holders (0.30 x 0.15 x 0.20 m) and were presented in the middle of a "real world viewing box" (1.0 x 1.0 x 0.8 m). The box was lined with multicoloured patches made of a variety of materials. Illumination of the scene was provided by a computer controlled, calibrated LCD-projector (Panasonic PT AE 1000E). The observers (n=5) viewed the samples frontally (viewing distance 0.90 m), with their heads fixed. All samples appeared achromatic after 2 min adaptation to the standard illumination (chromaticities of D65). Colour constancy was quantified by an achromatic setting method, after adaptation to one of four test-illuminants (selected from the cardinal axes in an equiluminant plane in colour space). The colour of all surfaces/objects was known, i.e. "white" under standard light. We found a strong influence of 3D shape on colour constancy performance: colour constancy was highest for the known 3D objects, followed by performance for the arbitrary 3D objects, and least for the flat surfaces. We conclude that the interaction between 3D-form and material provides local, object related cues for colour constancy. Object knowledge (object memory) seems to further support these processes.
Meeting abstract presented at VSS 2013