Research And Extension Of CAM16 Color Appearance Model

With the rapid development of science and technology,imaging and display devices are more and more popular.Color appearance model is an important connection space for cross-media color reproduction.The CAM16 proposed by Li et al.is a more widely used color appearance model,and therefore it has become a hot topic in the field of color science.The International Commission on Illumination（CIE）adopted the improvements proposed by Wang et al.and Gao et al.for the CAM16 model in 2022 and renamed the improved CAM16 as CIECAM16.This study was conducted in 2019 on issues related to the CAM16 color appearance model and focuses on the following:（1）.The domain and range for the CIECAM16 forward transformation are determined by linear programming approach.The domain and range for the conventional color appearance models CIECAM97s and CIECAM02,as well as the current CIECAM16forward transformation,have not been determined.It remains an open question,and causes many problems in the field of cross-media color reproduction.In this study,a linear programming approach is used with spectral reflectance as the optimization variable,a fixed Y value as the constraint,and minimizing and maximizing the X value as the optimization objective.Similarly,with fixed Y and X values as constraints,the maximum and minimum values of Z can be obtained,and thus the cubeΩ_F in XYZ space can be obtained.Then,the cubeΩ_R of color appearance attributes spaceJ,a_M,b_Mderived from the CIECAM16 color appearance model under the specified observation conditions can be calculated fromΩ_F.Finally,Ω_F andΩ_R are called the domain and range for the CIECAM16 forward transformation,respectively.（2）.A visual color difference dataset,COM（or more exactly COM-Corrected）was accumulated by CIE TC1-55,and evaluation of color difference prediction with CAM16-UCS using CIE 2°and 10°observers.CAM16-UCS is a uniform color space derived based on the CAM16 color appearance model,so the CAM16-UCS takes tristimulus values calculated using the CIE 2°observer as input.However,the tristimulus values for color difference are usually calculated using the CIE 10°observer,as in the CIEDE2000 or CIELAB formula.So,there was a debate as to whether CAM16-UCS can be used to predict color difference.Firstly,the reflectance of each pair of color samples in the visual color difference datasets:BFD-P,Leeds,RIT-Du Pont and Witt was generated based on the given tristimulus values,which were used as test data for the CAM16-UCS.Then,the predicted values of color difference（35）E_2,F and（35）E_10,F at 2°and 10°observers,as well as the visual color difference（35）V_{1 0},were evaluated by the CIE recommended model evaluation index STRESS values.Comparison results showed that the difference between（35）E_2,Fand（35）E_10,F was not statistically significant.Statistical tests showed that the differences between（35）E_2,F and（35）V₁₀,and between（35）E_10,F and（35）V₁₀ were not significant.Therefore,this study shows that CAM16-UCS can be applied to color difference calculations in the imaging,lighting,printing and packaging industries.（3）.Based on von Kries’modeling idea,the symmetry and transitivity of CAT16chromatic adaptation transforms are improved.In this study,we first analyzed chromatic adaptation transforms such as CAT02,CAT16,Fairchild and Smet,and found that most CATs don’t satisfy the symmetry and transitivity.Then,by changing the value of factor Y_W and the structure of von Kries adaptation factors in CAT16 model,the improved CAT16（Gv K model）has symmetry and transitivity while ensuring the model prediction accuracy.Finally,based on the Gv K model,the optimization of the formula for calculating the adaptation factor D was completed,and the optimization results led to a significant improvement in the prediction accuracy of the Gv K model.Note also that the Gv K model proposed in this study has been adopted by CIE in 2022 and applied in the CIECAM16.（4）.The experimental dataset of Fu et al.at Derby University was used to extend the CAM16 color appearance model for the prediction of unrelated color appearance properties.The conventional color appearance models CIECAM97s and CIECAM02,as well as the current CAM16 model,are unable to predict the color appearance properties of unrelated colors.A new color appearance model,CAM20u,for predicting the color appearance of unrelated colors,was developed based on the CAM16 color appearance model for related colors.The performance of the CAM20u model was tested using the three available datasets（Gent,Derby and ZJU）and it is shown that it outperforms other models for predicting brightness and colorfulness,but with no significant improvement in the prediction of absence-of-white.This indicates that,in principle,color appearance models for related colors can be extended to predict the appearance of unrelated colors.（5）.The experimental dataset of Xiao et al.at Leeds University was used to extend the CAM16 color appearance model for the size effect correction function.The perceived color appearance of an object varies with the size of that object or the viewing angle.The CAM16 color appearance model is only applicable to the tristimulus values under the CIE1931 2°standard observer as model input.Based on available visual size-effect data,it is found that the size effect can be compensated in the CAT16 cone-like space.Furthermore,it is found that,similar to the von Kries chromatic adaptation,size-effect correction can be operated independently on the different cone channels.Hence,the size effect can be easily built into the CAM16 model to predict the color appearance for stimuli of varying sizes.