提出了一种基于色域分析与聚类分析结合的基色筛选算法。该方法对传统4色色域和目标高保真色域进行分析，筛选出4色色域外的特征点，并对其光谱反射率进行主成分分析，确定基色的个数m。以m作为分类数，对集合中特征色光谱反射率进行聚类分析，在m个分类光谱中选择和特征色平均光谱距离最远的m个光谱作为基色光谱，在基色库中选择和m个光谱相关距离最小的实际专色作为最终基色。实验结果表明，提出的基色选择算法能够准确估计出新基色的光谱反射率，最大程度实现了目标高保真色域的再现。

To improve the color reproduction accuracy of image among digital devices in spectral color reproduction, a novel profile connection space is proposed. The profile connection space is composed of three sets of tristimulus determined by the principal components of typical illuminants and CIE1931XYZ standard colorimetric observer. The spectral reflectance of three common used atlases is transformed to three deformations of the profile connection space, and then transformed back to the spectral reflectance. The color difference between the original and transformed reflectance are very small for the common used illuminants, which indicates that the profile connection space is a good candidate for spectral color reproduction.