Brightness, Lightness, and Specifying Color in High-Dynamic-Range Scenes and Images

Traditional color spaces have been widely used in a variety of applications including digital color imaging, color image quality, and color management. These spaces, however, were designed for the domain of color stimuli typically encountered with reflecting objects and image displays of such objects. This means the domain of stimuli with luminance levels from slightly above zero to that of a perfect diffuse white (or display white point). This limits the applicability of such spaces to color problems in HDR imaging. This is caused by their hard intercepts at zero luminance/lightness and by their uncertain applicability for colors brighter than diffuse white. To address HDR applications, two new color spaces were recently proposed, hdr-CIELAB and hdr-IPT. They are based on replacing the power-function nonlinearities in CIELAB and IPT with more physiologically plausible hyperbolic functions optimized to most closely simulate the original color spaces in the diffuse reflecting color domain. This paper presents the formulation of the new models, evaluations using Munsell data in comparison with CIELAB, IPT, and CIECAM02, two sets of lightness-scaling data above diffuse white, and various possible formulations of hdr-CIELAB and hdr-IPT to predict the visual results.