Henry R. Kang

Neural network applications to the color scanner and printer calibrations

In the context of colorimetric matching, the intent of color scanner and printer calibrations is to characterize the device-dependent responses to the device-independent representations such as CIEXYZ or CIE 1976 L*a*b* (CIELAB). Usually, this is accomplished by a two-step process of gray balancing and a matrix transformation, using a transfer matrix obtained from multiple polynomial regression. Color calibrations, printer calibrations in particular, are highly nonlinear. Thus, a new technique, the neural network with the Cascade Correlation learning architecture, is employed for representing the map of device values to CIE standards. Neural networks are known for their capabilities to learn highly nonlinear relationships from presented examples. Excellent results are obtained using this particular neural net; in most training sets, the average color differences are about one ΔE ab . This approach is compared to the polynomial approximations ranging from a 3-term linear fit to a 14-term cubic equation. The results from training sets indicate that the neural net outperforms the polynomial approximation. However, the comparison is not made in the same ground and the generalizations, using the trained neural net to predict relationships it has not been trained with, are sometimes rather poor. Nevertheless, the neural network is a very promising tool for use in color calibrations and other color technologies in general.

Applications of color mixing models to electronic printing

The purpose of this study is to obtain some quantitative measures for the applicability of several color mixing models to a halftone printer. The printer, a Canon Color Laser Copier 500 (CLC-500), is treated as a black box and the measures are the difference between the calculated and measured spectra and Δ E ab . Well-known color mixing theories of the Neugebauer equations, Yule-Nielsen model (YN), Clapper-Yule multiple internal reflections (CY), Beer-Bouguerlaw (BB), and Kubelka-Munk theories (KM) are applied to CLC-500 to see how well they can fit the experimental data. Results indicate that the spectral 8-color Neugebauer model has marginal success in fitting the experimental data and the relaxed 3-color version does not fit the data weli. Both YN and CY approaches can fit the data within printer variability. The fittings are rather poor for BB and KM. By using the halftone correction factor, good agreements are obtained for the BB and single-constant KM.

Color scanner calibration of reflected samples

This paper presents a methodological approach for integrating non-colorimetric scanners with CIE standards as a means toward a device independent process. The calibration is aimed at reflected samples by employing photographic, thermal transfer, and xerographic prints. Correlations between scanner responses to CIE standards are established through a common test object using a two-step process of the gray balancing and the matrix transformation. A series of polynomials, ranging from a three-term linear combination to a twenty-term cubic equation, is used for converting device values to a CIE color space. The ability to fit colors that are not in the training set by a polynomial is examined. Results indicate that lower order polynomials fit colors equally well whether a color is in the training set or not, but the accuracy of interpolation decreases as the number of terms in the polynomial increases. We study the generality of this calibration method with respect to input materials. The transformation is material-dependent. Within the experimental uncertainty, however, there exists a unified transfer matrix for photographic materials and another one for paper substrates. Finally, we extended this method to deal with the mismatched illuminants for viewing and calibration. An empirical white point conversion method is proposed and tested; good approximations to the measured results are obtained when the interchange of illuminants occurs.© (1992) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Comparisons of three-dimensional interpolation techniques by simulations

Several 3D interpolation techniques for the color space transformation are compared via software simulations. Comparisons are made among four geometric interpolations, trilinear, prism, pyramid, and tetrahedral, with respect to the look-up table (LUT) size and packing. Three different LUT sizes and two ways of packing, uniform and nonuniform, are applied to the forward and inverse transformations of the XeroxRGB and CIELAB. Each simulation is tested by a set of 3072 points that are sampled around the entire RGB color space. Results indicate: (1) Interpolation errors of various 3D geometric interpolations are about the same and the errors with respect to the true values decrease as the LUT size increases. (2) The interpolation error peaks at the center of the cell and diminishes at nodes (lattice points). (3) The highest error occurs at the darkest region. For equally spaced LUTs, the error drops quickly as the level increases. (4) Nonuniform LUTs have a much lower fundamental error peak but the errors are rippled to the higher levels; this gives a more even error distribution and a better average value. From this study, it is conceivable that the colorimetric reproduction can be achieved to a very high degree of precision. With proper packing, the 3D interpolation provides the capability to closely approximate the true values in all regions of the color space. From the considerations of the implementation cost and computation speed, the tetrahedral interpolation is particularly attractive.

Gray component replacement using color mixing models

A new approach to the gray component replacement (GCR) has been developed. It employs the color mixing theory for modeling the spectral fit between the 3-color and 4-color prints. To achieve this goal, we first examine the accuracy of the models with respect to the experimental results by applying them to the prints made by a Canon Color Laser Copier-500 (CLC-500). An empirical halftone correction factor is used for improving the data fitting. Among the models tested, the halftone corrected Kubelka-Munk theory gives the closest fit, followed by the halftone corrected Beer-Bouguer law and the Yule-Neilsen approach. We then apply the halftone corrected BB law to GCR. The main feature of this GCR approach is based on the spectral measurements of the primary color step wedges and a software package implementing the color mixing model. The software determines the amount of the gray component to be removed, then adjusts each primary color until a good match of the peak wavelengths between the 3-color and 4-color spectra is obtained. Results indicate that the average (Delta) Eab between cmy and cmyk renditions of 64 color patches is 3.11 (Delta) Eab. Eighty-seven percent of the patches has (Delta) Eab less than 5 units. The advantage of this approach is its simplicity; there is no need for the black printer and under color addition. Because this approach is based on the spectral reproduction, it minimizes the metamerism.

Issues of Digital Color Imaging

Excerpt A digital color image is a multidimensional entity. It is sampled in a 2D plane with width and length, having quantized values in the third dimension to indicate the intensities of three or more channels (trichromatic or multispectral) for describing color attributes. The smallest unit in the 2D image plane is called the picture element (pixel or pel), constituted by the pixel size (width and length) and depth (the number of tone levels). However, the appearance of a color image is much more intriguing than a few physical measurements; there are psychological and psychophysical attributes that cannot be measured by any existing instruments—only by human vision. Therefore, the most important factor in color-image analysis is human vision because a human being is the ultimate judge of image quality. Human vision provides the fundamental guidance to digital-imaging design, analysis, and interpretation. Because human vision is subjective and instrumentation is objective, there is a need for establishing the correlation between them. This makes digital color imaging a very interesting and complex process, involving human vision, color appearance phenomena, imaging technology, device characteristics, and media properties. With rapid advances in color science and technology that resulted in a better understanding of color images in the forms of human visual system (HVS) and color appearance model (CAM) on the one hand, and the developments of various color-imaging technologies and computation approaches on the other, it is now possible to address the color-image reproduction at the system level in a quantitative manner and to produce good image quality across all media involved. To achieve these goals, cross-media color-image reproduction consists of four core elements: device characteristics and calibration, color appearance modeling, image processing, and gamut mapping. This chapter describes the scope and complexity of the digital color imaging at the system level and provides the sRGB⇆CIELAB conversions as examples to illustrate the modular architecture and technical requirements for a successful implementation.

Image Color Analysis

The sections in this article are 1 Basics of Image Color Analysis 2 Color Description and Formulation 3 Human Visual Responses 4 Color Appearance Model 5 Color Mixing Models 6 Circular Dot Overlap Model 7 Device-Independent Color Imaging

Microcluster line screens and frequency analyses.

Microcluster halftoning is a hybrid approach between clustered-dot and dispersed-dot ordered dithers. The concept and design principles of microcluster dots have been published elsewhere. This paper reports the frequency analyses of microcluster line screens. First, the Fourier transform for the frequency-domain analysis is briefly reviewed. Several line screens are proposed, ranging from 8 to 144 levels. At each level, three dot growth patterns of the conventional, interlace, and microcluster line screens are provided. Frequency analyses of these screens and the corresponding dispersed-dot are presented and compared. Results indicate that the line screens can be made into high frequency and high tone levels. Generally, they behave like a dispersed dot in the highlight region, a line screen in the midtone region, and an inverted dispersed dot in the shadow region.