The Influence of Monitor White Point Chromaticity on Color Appearance
Lidija Mandić , Maja Strgar Kurečić, Sanja Mahović, Darko Agić
University of Zagreb, Faculty of Graphic Art
Getaldićeva 2, 10 000 Zagreb, Croatia
Mandic@grf.hr

Abstract: In cross media reproduction, it is important to be able to predict the appearance of colors under a wide range of viewing conditions. This research focuses on comparision of printed images with CRT displays. If a simple corimetric match is made between a printed image and a CRT display, the perceived color in the image typically do not match. This is due to differences in viewing conditions between the two displays. Such differences include changes in luminance level, white point chromaticity, and surround relative luminance. The aim of this research was to collect and evaluate additional corresponding-colors data using pictorial images presented in well-controlled viewing conditions. In this work, the influence of monitor white point chromaticity on color appearance is presented.
Keywords: cross-media imaging, color appearance, monitor white point


1. Intoduction
Digital colour imaging is one of today's most exciting and fastest growing fields. At first sight, it might seem that the endless flexibility provided by having signals in digital form should make all the problems easily solvable. The complexity of the situation arises from the different input, monitoring, and output media involved, and the different objectives required for different applications. It is often difficult or impossible to successfully exchange digital color images among different types of systems. The use of well–established measurement methods, along with color-management applications and image file format standards, should allow the open interchange of digital images among systems (Giorgianni, 1998).

The reproduction of color images in an ever-widening array of media is challenging for a variety of reasons. That include physical limitations of devices, colorimetric calibration and characterization of devices, viewing condtitions, intent and preferences. If a simple colorimetric match is made between a printed image and a CRT display, the perceived colors in the image typically do not match. This is due to differences in viewing conditions between the two displays. Such differences include changes in luminance level, white point chromaticity and surround relative luminance.

Conventional CIE XYZ colorimetry is useful for specifying color appearance under agiven set of viewing conditions and for determining whether two colors will match in a viewing configuration. However, two colors, which are a corimetric match, can appear qiute different if viewed under different viewing condition (Mandić,2002). Therefore, in recent years, researchers have tried to develop more comprehensive color appearance models, which are able to predict color appearance accurately across a range of viewing conditions (Fairchild, 1998). Color appearance models were derived from the results of psyphysical experiments using simple scenes (e.g. single colrs against a variety of different viewing environment. To determine whether these models are useful for realistic color reproduction, their performance must be evaluated for complex images. The aim of this research was to collect and evaluate additional corresponding-colors data using pictorial presented in well-controlled viewing conditions. The two most commonly used media, CRT-displays (soft copy) and printed images (hard copy), were included in this research using four complex images. The viewing conditions varied in white point chromaticity of monitor.


2. Experiment
Four printed images containg pictorial information were used as the originals. One scene was a natural scene (grass and sky), the other contains saturated colors from the majority of hues (masks), third was skin and fourth one neutral scene (snow-neutral tones). These images were continous tone images printed on mat papers using a large format printer at 254 dpi. The original hardcopies were captured by an Agfa scanner at 72 dpi to provide RGB data for processing the CRT reproduction. The scanner and CRT display were calibrated and characterized using Macbeth software ProfileMakerPro 4.

The experiment was taken in dark room. Printed hard copies were illuninated and viewed under Just light source that simulated CIE Standard Illuminants D50. Reproduction was displayed on Mitsubishi monitor at 72 dpi and had the same physical size as the originals. The CRT white point was set to the chromaticity coordinates close close to 5000K (CIE Illuminant D50), 6500K (CIE Illuminant D65) and 9500K. Each soft copy or hard copy image had a 5 mm white border , which was the reference white for chromatic adaptation purposes.

A total of five observers, all experienced in using Adobe Photoshop, took part in the entire experiment. All the observers had normal color vision as evaluated by Ishihara plates and a Farnsworth-Munsell 100-hue test. Observers sat approximately 25´´ from the printed originals and CRT screen. The printed originals and CRT reproduction were placed 90° from each other with respect to observers in an L-shape arrangement. All experiments were carried out in a darkened room, so that only the print or CRT images occupied observer field of view. Observers adjusted the rendition of of the scene to match the color appearance of the print.

The CRT image adjustments were accomplished using Adobe Photoshop. Observers were allowed to use any of the color adjustment tools in Photoshop, but were not allowed to perform spatial manipulation of the images. The length of each experimental session was left to the discretion of the observers. Each experiment began with the same starting-point image, uncorrected scans, and the matching was perform gor all images with one monitor white point, then with another monitor white point. Every time the monitor was calibrated .

After observers completed the various matching tasks, the resulting images were saved for later processing. The images were segmented into several number of object region, depend of image context.

3. Results
The actual prints used in the experiment were measured using a Gretag Spectrolino 45/0 spectrophotometer (Hunt, 1991). Measurements were made on the prints by sistematically sampling the same image regions and then averaging the tristumulus values.

Results are presented in Tables 1 – 4, and test images and the region that were measured are shown in Fig. 1-4. After measuring, CIELAB metric lightness (ΔL), chroma (ΔL) and hue (ΔH) were, calculated between the printed images and images displayed on CRT. Color differences ΔE 94 and ΔE 00 (Luo, 2000) were calculated for each region of image, and for three different white point of monitor, D65, D50 and D95.

The reserch was concentrated on the relative appearance attributes of lightness, chroma and hue, because imaging systems are typically limited to three degrees of freedom in color reproduction. Lightness, chroma and hue represent the most intuative dimensions that observers typically use to describe complex colored stimuli. The results of these experiments show that some of the observers were producing matches more closely represented their preference rather than an accurate reproduction of the printed image.


Figure 1. Test image 1

Table 1. Data for image 1




Figure 2. Test image 2

Table 2. Data for image 2




Figure 3. Test image 3

Table 3.
Data for image 3




Figure 4. Test image 4

Table 4. Data for image 4


3. Conclusion
Psychophysical experiments were carried out to assess the influence of monitor white point on image appearance. The largest color differences were noticed in white and black regions, due to the difference of media. The smallest differences in chroma and lightness value were when matching was done with monitor white point D95. The accuracy of matching between hard and soft copy depend of image content. Our further research will be oriented to influence of other parameter and development of transformation that will give the optimal start point.

References
[1] E.Giorgiani, T.Madden,(1998) Digital Color Management, Addison-Wesley, Harlow
[2] M.D.Fairchild, (1998) Color Appearance Model, Addison-Wesley, Harlow
[3] L. Mandic, S. Grgic, T. Kos (2002) “Color Appearance Models”, 9 th International Workshop on Systems, Signals and Image Proccessing IWSSIP 2002, Manchester, UK, November,
[4] M.R.Luo, G.Cui, (2001) “The Development of the CIE 2000 Color-Difference Formula:CIEDE 2000, COLOR research and application, Vol. 26, No. 5, October 2001,pp 340-350
[5] R.W.G.Hunt, (1991) " Measuring Colour", Ellis Horwood, New York