Abstract
The previous studies have shown that, among many factors that influence color constancy of a scene, the binocular disparity of surfaces in the scene may be an effective one. Most studies have used 3D scenes with binocular disparities produced by two computer-generated stereoscopic images on a display. In this study, however, we used real objects viewed with a synopter to eliminate the binocular disparity of surfaces. This viewing condition allows observers to use a large number of depth cues except binocular disparity for understanding the 3D structure of the scene. In the present experiments the observer's task was to adjust the chromaticity of the test patch located in a real 3D scene so that it appeared as an achromatic surface. The surroundings consisted of painted geometric 3D objects arranged on stairs. An LC projector illuminated the surrounding objects with correlated color temperatures of, 3000 K, 6500 K and 20000 K. We tested three different viewing conditions, (1) binocular viewing condition (with normal disparity), (2) monocular viewing condition (with indefinable disparity), (3) synopter viewing condition (with zero disparity). The results showed that quite good color constancy was obtained in all viewing conditions, and that the binocular viewing condition yielded the best color constancy. These findings indicate that the lack of binocular disparity information reduced the degree of color constancy despite the existence of other depth cues sufficient to understand the 3D structure of the scene. This suggests that the binocular disparity contributes to color constancy by not just providing the object locations or 3D arrangements in a scene.