Munsell Color Science Laboratory Newsletter Winter 1994 Testing Chromatic - Adaptation

Testing Chromatic-Adaptation Models Under the guidance of Mark D. Fairchild, MCSL has been conducting a series of experiments to test the effectiveness of chromatic adaptation and color appearance models in many situations. For my Color Science M.S. thesis, I have focused on testing chromatic adaptation models with simple field, printed object colors. I examined the performance of the CIELAB, CIELUV, Fairchild ’91, Hunt ’91, LABHNU2, Nayatani et al. ’90 and von Kries models. Two versions of the Fairchild model were used, hardcopy and softcopy. Also, the Hunt model was tested in two versions, with and without discounting the illuminant. My research involved two phases: mathematical examination and psychophysical comparison. In the first phase, we used each model to calculate corresponding colors for 40 Munsell patches under a variety of illuminants and luminance levels. These corresponding color predictions were then compared model to model on a calibrated Sony Trinitron monitor. Very large differences between the models were readily apparent. Yet, we could not determine which set of model predictions corresponded most closely to actual color perception. We then designed a psychophysical experiment to compare the models’ corresponding color predictions under daylight to an original sample under incandescent illumination at a luminance level 1/10 that of the daylight source. This was accomplished using a new successive-ganzfeld haploscopic viewing technique, developed specifically for this research. Using a paired comparison design, twenty-six observers were asked to decide which of the two presented models’ corresponding color predictions was the best match to the original sample. Ten areas of color space were judged, including a neutral scale. We used zscore and logistic scale values to analyze the data. The Hunt model was found to predict the best matches for each color sample, and across all ten color centers. No other model’s effectiveness approached that of the Hunt model. However, two versions of the Hunt model were tested, and neither performed consistently better, although they both outperformed all other models. Dr. Hunt has reviewed these results and believes an alternative form of his model would provide performance consistency over a wide chromatic range, and yet still retain its shown effectiveness for object colors. As of this writing, none of the tested models have been proven to provide optimal prediction of corresponding object colors, but newly-revised models show some promise. Elizabeth Pirrotta