基于Sigmoid权函数的分段色彩匹配模型

为兼顾对高低浓度色彩再现精度的要求，解决单一模型进行色彩匹配时出现的“低浓度细节丢失、高浓度色彩饱和度低”的问题，提出一种基于单色样本学习的分段式色彩匹配方法．首先，根据模拟退火技术求解出最佳：分段阈值，将样本数据分成高、低浓度两类，分别对其样本数据进行学习，建立高、低浓度色彩匹配模型，然后，基于sigmoid权函数对两个模型学习得到的曲线进行合成．实验结果表明：与单一色彩匹配模型相比，这种分段式色彩匹配模型可以在不增加测试样本数量的前提下，有效提高色彩匹配尤其是在高、低浓度端的匹配精度；用sigmoid函数作为权函数的曲线合成方法，与一般常用的曲线合成方法相比，不仅可以保证高、低浓度色彩的匹配精度，而且还能有效处理分段式建模中两段曲线的光滑衔接问题，避免曲线衔接点处“颜色反转”现象的发生．

A stepwise color matching model based on sigmoid function

A stepwise color matching method by learning pure - color samples is proposed to improve the color reproduction accuracy of both high and low density colors. First, the pure - color sample data is partitioned into two clusters of samples according to the optimal threshold computed by a simulated annealing method. The final fitting curve is then linked up by computing the sigmoid weighted sum of the two curves that were obtained by learning the two clusters of samples. Experimental results indicate that the proposed method can improve the accuracy of color matching with no extra color samples that should be measured. Compared with the commonly used curve -composing method, this method based on Sigmoid function not only ensures the reproduction accuracy of both high and low density colors but also avoids the occurrence of abrupt change in color appearance in the joining point by joining the two piecewise curves very smoothly.