Abstract
The visual impression produced by a complex scene depends critically on the rendering light source. Some light sources are clearly preferred by observers for optimal visualization of specific objects. Thus, when observers were asked to choose the preferred correlated color temperate (CCT) of a daylight illuminant for rendering artistic paintings of naturalistic scenes they preferred a CCT close to 5000 K (Pinto et al, Journal of the Optical Society of America A, 25 (3), 623–630, 2008). The reason for this preference and whether it extends to other types of scenes are largely unknown. The goal of the present work was to investigate psychophysically the CCT preferred by observers for rendering natural scenes and to investigate computationally which image properties influence this preference. Twelve observers viewed images of 8 natural scenes, from urban and rural settings, displayed on a calibrated CRT monitor. The images were obtained from a hyperspectral database of natural scenes from which the spectral reflectance of each pixel of the scene could be derived (Foster et al, Visual Neuroscience, 23 (3–4), 341–349, 2006). The images were generated with daylight illuminants whose CCT could be adjusted by the observer over the range 3600–25,000 K. The task of the observer was to set the CCT for the “best visual appreciation” of the scene. Consistent with data for artistic paintings, the distribution of observers' preferences had a maximum at a CCT close 5000 K. In the computational analysis, the images were analyzed in CIELAB space as a function of the CCT of the illuminant. It was found that image-quality metrics, such as image variance and image entropy, expressed in terms of the achromatic and the chromatic attributes of the space, varied considerably across the CCT range. But for chromatic attributes these variations showed a point of inflection in the range 3000–5000 K. These results suggest that there is a common daylight color that gives optimum perceived image quality for both natural scenes and their artistic representations.