多墨打印系统的Neugebauer光谱特征化模型研究

发布时间：2018-01-03 08:32

本文关键词：多墨打印系统的Neugebauer光谱特征化模型研究　出处：《浙江大学》2012年硕士论文　论文类型：学位论文

【摘要】：随着光谱成像技术的发展,打印系统光谱特征化研究也逐渐成为相关领域的研究重点。重构目标色的光谱反射比以复现该颜色的方法具有诸多优点,如预测精度高、可避免同色异谱现象等。相对于传统的四墨打印机,多墨打印系统具有更大的色域、更多的候选配方组合,为光谱图像的精确复现提供了很好的硬件平台。因此,本文采用胞元式Yule-Nielsen(?)指数修正的光谱式Neugebauer(?)莫型(简称为CYNSN模型)对多墨打印系统的设备特征化研究具有重要的理论意义和应用价值。首先,本研究实验测定了12墨彩色打印系统HP Designjet Z3200ps的各项物理特性,包括墨水干燥时间、墨水稳定性、均匀性、重复性、阶调响应曲线、色域以及墨量限制等。然后,通过设计并制作训练样本建立了5个三墨和6个四墨正向CYNSN模型,同时对网点扩大模型和Yule-Nielsen n-facto进行了优化,最终确定参数设置,各正向模型的平均色差△Eoo(D65,10°)均小于或接近1。同时,基于线性回归迭代算法和J.Guo等人提出的胞元搜索算法,建立了四类反向CYNSN模型,均具有优异的可逆性,大部分模型的平均色差△E00(D65,10°)为0.2左右。以其中一类反向模型为例,通过打印预测配方测试了各反向CYNSN模型的实际精度,且达到了较好的复现效果,其平均同色异谱指数均小于0.7,但此类反向模型具有一定的系统误差。此外,比较了光谱式与三刺激值式CYNSN模型的精度,验证了光谱式模型具有更高的预测精度、且可以有效减少同色异谱现象。分色准则算法是一种如何从候选配方中选择最优配方的方法。针对所建立的11个CYNSN模型,本课题比较和分析了7种基于光谱匹配精度指标的分色准则算法。实验结果显示7种分色算法对打印样本的预测精度和实际精度较为一致,因此只需采用最简单的算法即最小△Eoo(D65,10°)即可。对于孟塞尔检验样本,结合一个色差和两个同色异谱指数的指标avgE00M1分色方法的表现最好。除了很少一部分低明度的打印样本和位于中高明度或彩度的黄色孟塞尔色样,avgE00MI对其它色区的绝大部分样本的实际精度都较高,其平均色差和同色异谱指数均在2附近。本课题提出了基于波长校准和主元分析法的两类正向和反向CYNSN模型标定算法。两类方法均通过新墨盒打印少量标定样本建立与原墨盒颜色特性之间的关系,实现了以现有特征化模型在新打印环境中的高精度预测。通过对检验样本标定精度的比较分析,基于主元分析法的PPM (printpredict model)对现有的正向CYNSN模型标定效果最好,相对于基于波长校准的算法分别下降了0.11～1.84个△E00单位(平均为0.62),有些情况下甚至高于原墨盒的原始精度,因为其对墨水颜色变化进行高精度建模,同时还补偿了部分现有正向CYNSN模型的系统误差。至于反向CYNSN模型,两类方法均达到了预期的效果,基于波长校准的算法精度相对更高,而基于主元分析法的方法在个别情况下对人眼敏感波段的标定精度尚不理想。最后,对本论文研究的主要内容和结论进行了总结,并对后继进一步的研究工作做出了展望。
[Abstract]:With the development of spectral imaging technology, the study of spectral characteristics of the system of printing has become a focus of research in related fields. The spectral reflectance reconstruction of target color than the color reproduction method has many advantages, such as high accuracy, can avoid the phenomenon of metamerism. Compared with the four traditional ink printer, printing system with multiple the larger color gamut, combination of the candidate formula more, provides a good hardware platform for the precise reproduction of spectral images. Therefore, the cell type Yule-Nielsen (?) Neugebauer spectrum correction index (?) Mo type (CYNSN model) to study on multiple device characteristic printing system has important the theoretical significance and application value.
First of all, the physical properties of 12 color ink printing system of HP Designjet Z3200ps were determined in this study, including the ink drying time, ink stability, uniformity, repeatability, tone response curve, color gamut and ink limit. Then, 5 three and 6 four ink ink forward model is established through CYNSN design and the production of training samples, while the dot gain model and Yule-Nielsen n-facto were optimized, and ultimately determine the parameter setting, the average color difference Delta Eoo forward model (D65,10 degree) were less than or close to 1. at the same time, the cell search algorithm based on linear regression and iterative algorithm proposed by J.Guo et al, established four kinds of reverse CYNSN model that has excellent reversibility, the average color difference Delta E00 most model (D65,10 degree) is about 0.2. A kind of reverse model as an example, through the actual test print prediction formula fine each reverse CYNSN model Degree, and to achieve a better reproduction, the average metamerism index were less than 0.7, but the reverse model has a certain system error. In addition, compared with the spectral type three stimulus values CYNSN the accuracy of the model, verify the spectral model has higher prediction accuracy, and can effectively reduce the same metamerism.
The separation criterion is a formula from the candidate selection method to the optimal formula. For the 11 CYNSN model, this paper compares and analyzes the 7 kinds of colour separation criterion based on spectral matching accuracy. The experimental results show that 7 kinds of printing color separation algorithm samples the prediction accuracy and the actual accuracy is consistent therefore only by the most simple algorithm which is the minimum Eoo (D65,10 DEG). The Munsell test samples, a combination of color and two metamerism index avgE00M1 color separation method is the best. In addition to a small part of low lightness print samples and in the clever degree or saturation of the Munsell color yellow like the actual accuracy of avgE00MI on the other color area most of the sample is higher, the average color difference and metamerism index were near 2.
This paper puts forward the calibration algorithm for wavelength calibration and the principal component analysis of the two types of forward and reverse CYNSN model based on two methods. Through the new cartridge to print a small set of calibration samples and the relationship between the original ink color characteristics, achieve high prediction accuracy in the new printing environment using the existing characteristic model. Based on the test sample calibration precision, comparative analysis, principal component analysis method based on PPM (printpredict model) to forward the existing CYNSN model calibration effect is best, relative to the wavelength calibration algorithm based on decreased by 0.11 to 1.84 E00 units (average 0.62), the original accuracy in some cases even higher than the original cartridge, because its high precision modeling of ink color change, but also the system error compensation part of the existing forward CYNSN model. As for the reverse CYNSN model, two kinds of methods to achieve the desired effect, radical The accuracy of the Yu Bo length calibration algorithm is relatively high, while the method based on the principal component analysis is not ideal for the calibration accuracy of the human eye sensitive band in some cases.
Finally, the main contents and conclusions of this paper are summarized, and the future further research work is prospected.

【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP334.8

【参考文献】