打印机外文翻译奥普蒂化的光谱纽格伯尔模型打印机特性.doc

1、奥普蒂化的光谱纽格伯尔模型打印机特性摘要一个比色打印机需以其注入的一组油墨，预测由此产生的印刷颜色，通过反射或刺激值指定。该纽格伯尔模式已被广泛用于预测的半色调彩色打印机色反应。本文是介绍和比较优化纽格伯尔模型的技术。这些措施包括尼尔森因素的优化，这归因于光的散射及估计该点区域并扩展到精确位置的功能。一种新的技术描述了优化纽格伯尔初选使用加权谱回归。实验结果提出了它使用两个半色调屏幕：随机静电打印机或旋转点，和互联网上点屏幕。使用尤尔一尼尔森因子，细微的的框架，并使谱回归模型的准确度大大提高。1引言 比色打印机特性要求是建立在信号的输入到打印机和由此产生的印刷颜色的色度测量的关系上的。这种关系

2、，我们称之为打印机特性的功能，往往是得到了印刷和测量补丁的色彩大量运用了一些在已知的样本的测量插值。另一种方法近似于用打印机型号的特征函数。试图模拟打印机经常会产生有价值的直觉的基本物理过程，它在光，色彩和纸之间产生复杂的相互作用。模型的实际优点是打印机可以实现定性并相对较少的测量。打印机的模型精度显然取决于模型中的假设是否正确。该纽格伯尔模型已广泛用于二进制模式的彩色打印机并可采用各种色调屏幕。原始模型实质上是Murray-Davies equation（用于预测灰度反射）的色彩方法。打印机字的颜色由红绿蓝（RGB）三色组合，预计作为对固体套印的RGB值加权平均印刷；青色，品红和黄色的印刷，

3、由地区覆盖范围相对点的权重决定。在这里，R，G，B可能被认为是从印刷补丁反射光，通过3个已知光谱灵敏度的职能通过宽带过滤器，超过估计所有在五个可见光波长范围内的方法，从该地区的覆盖范围数字输入色调，包括一些直接测量和计算的结合。请注意，尽管这种模式提供设备值（CMY）的特征函数（三原色）色值，它是从色值到设备空间，最终是为了寻求色彩再现逆映射。几位研究人员讨论了反相的纽格伯尔方程。这个反演问题是一个非线性问题，可以解决在三个输入着色剂的情形C，M，Y，通过迭代方法。当着色剂数量超过3，问题就变得病态数着色剂组合可以导致生成相同的测量颜色。在这项工作中，只有前进定性问题的审查。笔者认为，一旦提出

4、一个精确的模型，样品的高密度，导出等多层面的拟合和插值技术可以很容易被用来颠倒模式。 各种方法已经被提出了提高原纽格伯尔方程的准确性。一个重要的现象，不占原模型，光散射的文件内。Yule and Nielsen 通过模型中的一个额外的参数方程纽格伯尔这种效果。其他研究人员提出了在之间的透射模式和一纸点扩散函数（PSF）的回旋空间光散射的形式更复杂的模型。尼尔森参数不明确的空间变化之间的相互作用光纸帐户，有人建议由鲁克德舍尔和豪泽，一个简单的关系可能与这个简单的模型和涤纶短纤存在的模型。维贾诺表明，而不是在纽格伯尔方程形式的收益率更高的精度宽带频谱。这一发现证实了作者。 Engeldrum表明，

5、无论是纸张和圆点的颜色相对点区域覆盖范围内的功能，并延长了纽格伯尔模型交代。利用该纽格伯尔方程的几何解释，开发了蜂窝Hueberger卢哥贝尔模式，即除固体套印颜色来计算最终输出的颜色。蜂窝技术将被视为在本文件。从德米歇尔的假设出发点区域覆盖范围内，并用序列二次规划方法来估计这些参数。探讨了波长依赖尤尔使用尼尔森的因素。 在这项工作中采取的做法是，以符合理想打印机的纽格伯尔模型的基本形式，并提供优化模型的参数，以便更好地适应实际打印机的特性的数学技术。因此，该模型简单保留，而其精度提高。优化的圣诞-尼尔森因素，技术点区域功能，细胞模型已被提交会议的出版物上发表。本文介绍这些技术在一个比较一致的

6、框架。此外，由在最近的一次会议报告的作者提出谱回归的新方法，更详细。实验结果提出了两种类型的半色调屏幕：随机或旋转点，和互联网上点屏幕上。 2纽格伯尔混合模型。二进制打印机，一个多种颜色的渲染，是通过不同的组合达到点着色剂的主要地区覆盖范围。在纽格伯尔方程预测从任意作为一种已知的基础上为2n的色彩，光谱反射加权平均印刷补丁的平均光谱反射率，其中N是系统着色剂数量。在这项工作中的N = 3 CMY打印机案件将用来描述各种模型。方程很容易推广到N着色剂，而事实上，青色，洋红色，黄色，黑色（CMYK）打印机将在实验结果中。对于CMY打印机，有8个的基础上的颜色，这是白皮书（宽）和1，2和3纯色覆盖（

7、即0和C，男，y的100的组合），这些基础色将称为纽格伯尔初选。在原来的纽格伯尔方程，预测的颜色是由三个宽带代表短期，中期反射和电磁波谱的长波长部分指定。在这项工作中，窄带反射光谱住宅（十）使用，而不是他们的宽带同行，因为前者一般收益率更高的准确性。3结论 本文几个打印机的纽格伯尔方程的模型已被描述和比较。一种新的技术已提出了优化光谱纽格伯尔初选使用最小二乘回归。提供各种型号的复杂性和精确度之间不同的权衡。表1总结了在测量数量而言，这些权衡需要制定出一个给定的模式和结果的准确性。一般来说，模型的准确度是成正比的测量数目，但人们可以方便地识别报酬递减的情况。产生的巨大收益来自使用尤尔一尼尔森改正

8、，并可以在细胞框架或全球谱回归模型。测量的数目增加的细胞模型与细胞数量成倍的要求，同时在准确增益减小。在34 = 81初选的情况常常是一个可以接受的折衷。在当地，在计算成本的大幅增加加权回归结果（未显示在表1），在没有获得准确对全球回归。 在执行回归的数学解释是，它提供了一个自由的模式（即初选额外的程度），因此适合优越的经验数据。从初选的光谱，如无花果，阴谋的研究。 10和11，已经从没有测量，观察到的趋势回归职能。因此，很难物理意义重视对这些阴谋，除了指出，在选定的样本良好的线性回归可能会导致的光谱测量噪声平均，从而得到更强大的预测。这项工作提出了一些进一步的扩展。谱回归应用于蜂窝架构可能会

9、导致没有额外的提高测量精度。阿谱初选的联合优化，点区域功能和尤尔一尼尔森因素可能导致对其中n因素是单独优化目前的做法更好的结果。这是身体有理由认为在光散射波长基板依赖，因此一个谱回归模型，支持n是X的函数可能会导致更准确的模型的扩展。然而，重要的是铭记之间的物理模型的复杂性和准确性实现权衡始终牢记实际打印系统。最后，这将是值得研究的印刷技术比其他如静电复印，喷墨机，胶印机等更具有效性。6Opti ization of the spectral Neugebauer model for printer characterizationAbstractA colorimetric printer

10、model takes as its input a set of ink values and predicts the resulting printed color, as specified by re?flectance or tristimulus values. The Neugebauer model has been widely used to predict the colorimetric response of halftone color printers. In this paper, techniques for optimizing the Neugebaue

11、r model are presented and compared. These include optimization of the Yule-Nielsen factor that accounts for light scattering in the pa?per, estimation of the dot area functions, and extension to a cellular model. A new technique is described for optimizing the Neugebauer primaries using weighted spe

12、ctral regression. Experimental results are presented for xerographic printers using two halftone screens: the random or rotated dot, and the dot-on-dot screen. Use of the Yule-Nielsen factor, the cellular framework, and spectral regression considerably increase model accuracy. ? 1999 SPIE and 1S&T.

13、1 IntroductionColor printer characterization requires that a relationship be established between the input signals to the printer and the colorimetric measurements of the resulting printed colors. This relationship, which we call the printer characterization function, is often obtained by printing a

14、nd measuring a large number of color patches and applying some interpo?lation among the measurements at the known samples. An alternative approach is to approximate the characterization function with a printer model. An attempt to model the printer often yields valuable intuition about the underlyin

15、g physical process, which constitutes complex interactions between the light, colorant, and paper. A practical advan?tage of modeling is that the printer characterization can be achieved with a relatively small number of measurements. The accuracy of the printer model clearly depends on the validity

16、 of the assumptions in the model.The Neugebauer model has been widely used to model binary color printers employing various halftone screens. The original model is essentially an extension of the Murray-Davies equation2 (used for predicting grayscale re?flectance) to the color case. The color of a p

17、rint, in red?green-blue (RGB) coordinates, is predicted as a weighted average of the RGB values of the solid overprints of the three printing primaries, cyan, magenta, and yellow (C, M, Y), where the weights are determined by the relative dot area coverages c, m, y constituting the print. Here, R, G

18、, B may be thought of as the reflected light from the printed patch, passed through three broadband filters with known spectral sensitivity functions, and summed over all wave?lengths in the visible range V. Methods for estimating dot area coverages from the digital input values to the halftone incl

19、ude some combination of direct measurement and cal?culation.Note that while the model provides the characterization function from device values (CMY) to colorimetric values (RGB), it is the inverse mapping from colorimetric to de?vice space that is ultimately sought for color reproduction. Several r

20、esearchers have addressed the problem of invert?ing the Neugebauer equations.3-6 The inversion is a nonlin?ear problem that can be solved, in the case of three input colorants C, M, Y, by iterative approaches. When the num?ber of colorants exceeds three, the problem becomes ill-posed, as several col

21、orant combinations can result in the same measured color. In this work, only the forward char?acterization problem is considered. It is the authors opin?ion that once a sufficiently dense sampling of an accurate forward model is derived, other multidimensional fitting and interpolation techniques7?8

22、 can easily be used to invert the model.Various methods have been proposed for improving the accuracy of the original Neugebauer equations. An impor?tant phenomenon not accounted for in the original model is that of light scattering within the paper. Yule and Nielsen9 modeled this effect via an addi

23、tional parameter in the Neu?gebauer equations. Other researchers10-12 have proposed more sophisticated models for light scattering in the form of a spatial convolution between the transmittance pattern and a paper point spread function (PSF). While the Yule-Nielsen parameter does not explicitly acco

24、unt for spatially varying interactions between light and paper, it has been suggested by Ruckdeschel and Hauser13 that a simple rela?tionship may exist between this simple model and the PSF based models. Viggiano14,15 demonstrated that working with the spectral rather than the broadband form of the

25、Neugebauer equations yields higher accuracy. This finding has been confirmed by the author.16 Engeldrum et a/.17 .I showed that the colors of both the paper and the dots are functions of the relative dot area coverages, and extended the Neugebauer model to account for this. Utilizing the geometric i

26、nterpretation of the Neugebauer equations, Hueberger19 developed the cellular Neugebauer model, where colors in addition to the solid overprints are used to calculate the final output color. The cellular technique will be considered in this paper. Lee et al.20 departed from the Demichel assumptions

27、on dot area coverages, and used a sequential quadratic programming method to estimate these parameters. Hua and Huang21 and Iino and Bems22. ex?plored the use of a wavelength-dependent YuleNielsen factor.The approach taken in this work is to conform to the basic form of the Neugebauer model for an i

28、deal printer, and to offer mathematical techniques for optimizing the parameters of the model to better fit the characteristics of real printers. Thus the simplicity of the model is retained, while its accuracy is improved. Techniques for optimizing the YuleNielsen factor, dot area functions, and ce

29、llular model have been published in previous conference publica?tions by the author.1624 In this paper, these techniques are presented and compared in a coherent framework. In addi?tion, a new method of spectral regression proposed by the author in a recent conference report25 is presented in greate

30、r detail. Experimental results are presented for two types of halftone screens: the random or rotated dot, and the dot-on?dot screen.The paper is organized as follows. In Sec. 2, the Neuge?bauer mixing model is briefly introduced, and is applied to the two types of halftone configurations. The light

31、 scatter?ing problem and YuleNielsen correction are described in Sec. 3. Methods for calculating the dot area functions are described in Sec. 4, and extensions to the cellular frame?work are presented in Sec. 5. A spectral regression tech?nique is described in Sec. 6, where the Neugebauer prima?ries

32、 are themselves treated as parameters to be optimized. Experimental results are presented in Sec. 7, and conclud?ing remarks are collected in Sec. 8.2 Neugebauer Mixing Modelor a binary printer, the rendering of a multitude of colors is achieved through combinations of varying dot area cov?erages of

33、 the primary colorants. The Neugebauer equations predict the average spectral reflectance from an arbitrary printed patch as a weighted average of the spectral reflectances of 2N known basis colors, where N is the num?ber of system colorants. In this work, the case of N=3 for CMY printers will be us

34、ed to describe the various models. The equations are easily generalized to N colorants; and indeed, cyanmagentayellowblack (CMYK) printers will be included in the experimental results.2.3 Effect of Screen DesignOne of the fundamental factors that affects model accuracy is the frequency of the halfto

35、ne screen.28 The ideal model assumes that the dots are rectangular in cross section, and hence result in a binary absorption profile as a function of spatial location. In reality, dots have soft transitions from regions with full ink to regions with no ink. If the halftone screen frequency is relati

36、vely low, or a clustered dot is used, the relative area of the paper covered by the transition pected to be relatively accurate. On the other hand, if the screen frequency is high, or a dispersed dot is used (as is the case with a Bayer screen or error diffusion), then a large fraction of the paper

37、is covered by transitory regions, and the model breaks down. The corrections discussed in Sec. 3 partially account for soft transitions; nevertheless, the reliability of the model has been seen to be greatest with clus?tered dot screens in the range of 300-400 halftone dots per inch.8 ConclusionsIn

38、this paper, several printer models based on the Neuge?bauer equations have been described and compared. A new technique has been proposed for optimizing the spectral Neugebauer primaries using least squares regression. The various models offer different tradeoffs between complex?ity and accuracy. Ta

39、ble 1 summarizes these tradeoffs in terms of the number of measurements required to derive a given model and the resulting accuracy. Generally, model accuracy is proportional to the number of measurements; however one can easily identify cases of diminishing re?turns. The significant gains come from

40、 using the Yule-Nielsen correction, and either the cellular framework or the global spectral regression model. The number of measure?ments required in the cellular model increases exponen?tially with the number of cells, while the gain in accuracy diminishes. The case of 34=81 primaries is often an

41、ac?ceptable tradeoff. The locally weighted regression results in a significant increase in computational cost (not shown in Table 1), with little gain in accuracy over the global regres?sion.The mathematical interpretation of performing regres?sion is that it affords an extra degree of freedom in th

42、e model (i.e., the primaries), and hence a superior fit to em?pirical data. From a study of plots of the spectral primaries, such as Figs. 10 and 11, no trend has been observed from measured to regressed functions. Hence it is difficult to attach physical significance to these plots, other than to p

43、oint out that linear regression over a well chosen sample set will likely result in an averaging of the spectral mea?surement noise, and hence a more robust prediction.This work suggests several further extensions. Spectral regression applied to the cellular framework may lead to improvement in mode

44、l accuracy with no additional mea?surements. A joint optimization of the spectral primaries, dot area functions, and Yule-Nielsen factor may lead to superior results over the current approach where the n fac?tor is optimized separately. It is physically plausible that the scattering of light within

45、the substrate is wavelength dependent; hence an extension of the spectral regression model that allows n to be a function of X may lead to a more accurate model. However, it is important to always bear in mind the tradeoff between the complexity of a physical model and the accuracy achievable in a practical printing system. Finally, it would be worthwhile to examine the validity of these models for printing technologies other than xerography, e.g., inkjet, offset press, etc. .此处忽略！