胞元纽介堡模型颜色预测性能的影响因素分析

目的 以色彩管理系统的研发为目标,研究多个因素对正向纽介堡模型颜色预测性能的影响,为模型应用提供参数优化方案.方法 借助Matlab平台模拟基于不同胞元等级、不同检验样本的位置和数量,以及不同胞元个性化修正方案的胞元纽介堡模型,并通过样本打印和测量实验,评价上述各因素对模型颜色预测性能的影响,确定模型的最优参数方案.结果 模型精度随胞元划分等级的增加而提升,但在4或5级胞元后精度逐渐稳定.另外,胞元内采样点数量、位置以及是否采用胞元个性化修正方案,对颜色转换精度都没有产生有效影响.确定将5级胞元划分,胞元中心采样和所有胞元统一修正指数作为胞元纽介堡模型的参数优化方案,与同类型算法模型的比较结果表明,在采样数量一致的情况下,参数优化的胞元纽介堡模型与i1 Profiler软件,胞元神经网络模型的预测色差都小于1个CIEDE2000色差单位,且相互间的差别都小于0.15,在系统误差范围内,胞元顶点距离插值算法的预测色差则达到了3以上.从算法结构上考虑,神经网络模型需要对所有胞元进行训练建模,计算量较大.结论 综合考虑算法精度和效率,参数优化的胞元纽介堡模型可以满足目前印刷工业中色彩复制的需要.

Influencing Factors of Color Prediction of Cellular Neugebauer Model

In the color management system, the color prediction performance of the forward color conversion model directly affects the color separation accuracy of the reverse model. Aiming at the research and development of color management system, this paper aims to study and analyze the several influencing factors on the color prediction performance of forward Neugebauer model, and provides parameter optimization scheme for the application of the model. With the help of MATLAB platform, the cellular Neugebauer models based on different cell-levels, different locations and quantities of test samples, and the different cellular correction schemes were simulated. Through the printing and measurement experiments of samples, the influence of above factors on the color prediction performance of the cellular Neugebauer model was evaluated, and then the optimal parameter scheme of the model was determined. The result of the experiment indicated that accuracy of the model increased with the increase of cell level, but no longer with significant changes at cell-level 4 or 5. In addition, the number and location of sampling points within the cell as well as the cellular correction scheme exerted no effective influence on the accuracy of color conversion. Based on the above analysis, this paper determined to use the five-levels cellular division, the cell-center sampling and the unified cellular correction scheme (all the cells share one correction index) to optimize the cellular Neugebauer model. Compared with the same type of algorithm model, and in the case of the same number of test samples, the predicted color difference of the cellular Neugebauer model with optimized parameters, i1 Profiler software and the cellular neural network model were all less than 1 CIEDE2000 color difference unit, and the difference among them was all less than 0.15, which was within the range of system error, and the predicted color difference of the distance-weighted interpolation algorithm reached more than 3. Considering the algorithm structure, the neural network model requires training and modeling of all cells, with a large amount of computation. Therefore, considering the accuracy and efficiency of the algorithm, the cellular Neugebauer model with optimized parameters can meet the demands of color reproduction in the current printing industry.