基于高质量对应色数据集的个性化色适应变换模型的研究

Researches on Individual Chromatic Adaptation Transforms Model Based on High-Quality Individual Corresponding Color Data

目前个性化颜色视觉的研究越来越受到重视,为推进个性化色适应变换模型的研究和应用,本研究进行了高质量个性化对应色数据的采集,并基于此推导了个性化色适应变换模型。首先通过颜色记忆能力测试筛选观察者,然后借助于图形用户界面(GUI)在两台显示器上完成对应色采集实验,观察者可以对颜色进行调节,实验重复3次;并基于此数据集用保持白点的光谱锐化算法推导出了个性化色适应变换模型的转换矩阵,并对模型进行了检验。对应色数据集的观察者自身重复精度为3.58ΔE_(00),观察者之间标准偏差为3.81ΔE_(00)。利用前两轮实验数据推导的色适应变换模型预测第三轮实验数据,平均色差为5.46ΔE_(00)。实验结果显示颜色记忆误差是影响对应色数据集精度的一个不可避免的因素;对应色采集实验中存在H-K效应,目前的色适应变换模型包括CAT02,不能预测色适应过程中存在的H-K效应,需要改进;保持白点的光谱锐化色适应变换模型推导算法的精度有待提高。

The research interest on the individual color vision is becoming more and more intensive.In this study,in order to promote the research and application of personalized color adaptation transformation model,the high-quality individual corresponding colors data were collected and individual chromatic adaptation transforms(CATs)were derived.Firstly,the observers were chosen by color memory ability test.Then the corresponding colors collecting experiments were executed on two monitors with graphic user interfaces(GUI).The observers could adjust reference colors on the GUI and the experiments were repeated by three times.Based on the data,individual CATs were derived by white-point preserving data-based spectral sharpening method and evaluated.The average intra-observer deviation of the data was 3.58ΔE_(00),and the average inter-observer deviation was 3.81ΔE_(00).The CATs derived from the first two rounds of the experimental data were adopted to predict the experimental data of the third round,and the average color difference was 5.46ΔE_(00).Some conclusions were obtained.The color memory error is an unavoidable factor that affects the precision and accuracy of the corresponding color data.The H-K effect works during the chromatic adaptation.The current chromatic adaptation transforms including CAT02 cannot predict H-K effect existing in the chromatic adaptation process and need to be improved.The precision of the white-point preserving data-based spectral sharpening method needs to be improved.