Nonlinear tone scale operations applied to RGB image data are important in many image processing applications. While such operations are intended to produce desirable changes in image luminance contrast and colorfulness, they may also introduce unwanted hue shifts of image colors. The magnitudes and distributions of these hue shifts in color space are affected significantly by the selection of the RGB primaries for the color encoding. This work has systematically investigated such hue shifts resulting from the application of typical nonlinear image processing transformations. Tests were performed using a highlight-to-shadow series for 18 different starting colors, representative of color transitions that occur in real scenes and images. Hue-shift metrics were calculated in four “perceptually uniform” color spaces: CIELAB, CIECAM97s, IPT, and OSA_UCS. Three sets of primaries were evaluated in this study, corresponding to those defined for the ROMM RGB, sRGB, and Adobe Photoshop Wide Gamut RGB color encodings. Of these, it was found that the ROMM RGB primaries introduced the smallest overall hue shifts, while the Wide Gamut RGB primaries introduced the largest overall hue shifts. These results were consistent across all of the uniform color spaces in which the hue shifts were evaluated.
Geoff Woolfe, Kevin Spaulding, Edward Giorgianni, "Hue Preservation in Rendering Operations – An Evaluation of RGB Color Encodings" in Proc. IS&T 10th Color and Imaging Conf., 2002, pp 317 - 324, https://doi.org/10.2352/CIC.2002.10.1.art00059