ABSTRACT
The color predictive accuracy evaluation of color appearance models is that some
predicted experiments are designed to evaluate the color reproduction ability of the
color appearance models, in which the subjective and objective data such as various
color appearance data, corresponding color data, color system data sets and chromatic
aberration data sets are used. The color accurate prediction of color appearance model is
an important factor whether the color appearance models can be widely used in color
research and industrial fields. Since the 1990s when color appearance model was
proposed, the testing color accurate evaluated experiments about the color prediction
accuracy of the color appearance models are the bases of the correction of the color
appearance models and a new color appearance model proposed. The color forecast
accuracy evaluation of the color appearance models has a theoretical significance to the
practical application of the color appearance model and the improving of color
appearance model.
In this paper according to their visual uniformity, Munsell new standard system
datasets and Chinese color system datasets chosen, pattern recognition evaluation
method is used to evaluate the color forecast accuracy of the eight color appearance
models (CIELAB, LLAB, RLAB, ZLAB, Hunt94, Nayatani, CIECAM97s, CAM02)
from the three aspects of the relative attributes of hue, lightness and saturation. The
results of the analysis is that different color appearance models have different forecast
superiority both in three color appearance relative attributes and in different color areas.
Principal component analysis theory is proposed for the color appearance model
prediction, and comprehensive evaluation index is used to make a comprehensive
ranking for the color predictive ability of the color appearance models.
Then 682 Chinese color samples selected, pattern recognition technology are used
to evaluate the eight color appearance model prediction of lightness, chroma, and hue.
There are two results that no color appearance model is good at predicting the three
attributes, and different color appearance models have the different predicted
advantages in different color areas. About the non-directional prediction results of the
color appearance model got from pattern recognition technology, a comprehensive
evaluation index is extracted from the four indicators of pattern recognition using
principal component analysis; in the ten main hue areas, the color appearance models
are made a comprehensive ranking; thus the view of color appearance model partition