DISPLAY HARDWARE AND SOFTWARE
Evaluate the Visual Performance of Your Display Display Calibration
This page discusses factors that affect colors on common kinds of displays in general terms. This is too large a topic for detailed discussion in this website. More detailed treatments are widely available. bibliography reference sources.
Characteristics of the users' display hardware and software interact with the users' perceptual properties to determine the appearance of colors. Display color gamut, gamma, and calibration need to be considered as constraints on the design of color graphics.
If the users you are designing for will all be using the same display hardware and software and it is well-maintained, you can design for that known system with a fair degree of confidence that most users will be viewing the intended physical colors. If a variety of systems will be used or the displays are poorly maintained you may need to design with wider margins of safety--higher luminance contrasts for symbols, wider chromatic separations of colors that must be discriminated, etc.
Among the most important determinants of the color gamut of a display are the chromaticity coordinates of its red, green, and blue primaries and its white point (white alignment). In recent years consumer desktop displays have begun to converge on some standard parameters.
Chromaticities of Monitor Primaries. Common displays have three primaries (red, green, and blue) with chromaticities falling within fairly small regions of the CIE diagram. This figure shows several that we have used and measured over the past few years. The SONY CRT has primaries close to those of the ITU-R BT.709 standard for HDTV displays. The red and green chromaticities of the LCDs are close to those of the SONY CRT, but the blue primaries of the LCDs tend to be shifted slightly toward green (higher y) relative to the CRT.
While not all of these chromaticity differences are insignificant, they are also probably not a major concern in color designs for most applications. On all of the displays shown here any colors used for labeling that are highly discriminable on one display should also be distinct on another. Any color which the user would be satisfied to describe with a one-or-two-word color name on one display should be well-described by the same name on another.
In practical terms,
the chromaticities of the display primaries must be considered fixed.
CRTs the primaries are determined by the phosphors, though some leakage
among the primaries can occur. On high-quality displays this is usually
minor. For LCDs the chromaticities of the primaries are jointly determined
by the fixed spectral properties of the color filters, the active elements,
and the backlight.
White Alignment. Another parameter that affects the color gamut in 3D color space is the color of the "white" that is displayed when the data inputs to all three primaries are at the maximum. The chromaticity of this white is called the white alignment. Most commercial displays are either adjusted at the factory to one of several internationally standard white colors or offer a means to select from these preset standard colors. "6500K", an average daylight white, is frequently used. The LCD displays we have measured all have either a fixed alignment to 6500K or offer it as a choice in the display controls. Another fairly common standard white alignment is 9300K, a bluer white than 6500K. As with other parameters the factory white-alignment standards of individual displays will vary within manufacturing tolerances.
The same digital input data will correspond to different color appearances on monitors with different white alignments in a side-by-side comparison. On a display adjusted to 9300K the luminance of the blue primary is higher than of a display adjusted to 6500K, so many colors will appear slightly bluer. The physical differences are least for colors that are almost entirely one primary or another and larger for colors with nearly equal contributions from two or more primaries.
These differences are likely to be larger than those due to slight variations of the display primaries. Even so, they may not be a major concern for the color designer for applications in which all of the displays have the same white alignment. If all of the user's work is on the same display, adaptation in the user's visual system will limit the effect of the differences on naming, discrimination, and luminance contrasts.
The transfer functions (sometimes called the display's "gamma function") of the display's primaries are a major determinant of the appearance of its displayed colors. These are the functions relating the digital input data values to the corresponding luminances of the primaries. In a perfectly designed and adjusted display the functions for the three primaries will be identical. This is a necessary condition for the chromaticity of gray to be the same at all luminances. The transfer functions will generally vary by design among display types. They will also vary among individual displays. The transfer functions for CRTs are largely determined by the physics of the tube, though some of the internal and external adjustments, e.g., the black level and contrast, affect part of the shape. The functions for LCDs are less dictated by the physics and could theoretically be designed to be very different from CRTs. Some LCDs have controls for selecting among several pre-programmed gamma functions. For several practical reasons most LCDs have been designed to have functions similar to those for CRTs. One is compatibility with existing electronic documents that were designed to display well on CRTs. There are also visual advantages, discussed below.
For precise color work the individual display's transfer functions must be measured with a photometer, and factors such as reflected light in the viewing situation must be taken into account.
When each curve is normalized to its maximum luminance the curves of the primaries can be seen to have nearly identical shape, a positively accelerated curve with a small "dark light" at zero. (Since the dark light is identical for the three primaries and the peak luminances are different, the normalized dark lights are different.)
Visual Consequences of the Transfer Function. The function relating perceptual brightness to luminance is negatively accelerated. Under some viewing conditions it has a roughly cube root form and in others it is nearly logarithmic. In either case, a constant luminance difference corresponds to a larger brightness difference at low luminances than at higher luminances. While the positively accelerated transfer function of CRTs originated in the physics of the device, it has the visual benefit of compensating somewhat for the nonlinearity of brightness perception. In imaging applications we would usually like to be able to perceive small differences in the image throughout the luminance range of the image. To achieve this we need a constant data difference to produce larger luminance differences at high luminances than at low, to compensate for the visual system's nonlinearity. The positively accelerated transfer function of the display partially satisfies this need.
With this figure you can evaluate how well you perceive small data differences in various parts of the grayscale of your display. Each cell of the figure is split into halves. One half has the digital data value labeled on the abscissa. The other half has digital data equal to Data minus Delta Data. Thus the pairs in the bottom row differ by one digital count and those in the top row differ by 10 counts. If you look carefully at a good display you should be able to barely see the dividing edge between the two halves in a few of the cells of the second row from the bottom. On my Cinema Display under the viewing conditions in my office I can see a difference across all of row 6, but differences of 5 or less are invisible in the 15 (leftmost) column and none are clear in the bottom row.
The conventional approach to color calibration of a display involves measurement of a number of parameters of its light output using spectrophotometric and photometric instruments. Which measurements and the required instrumentation depend on the desired accuracy. While several alternative approaches have been described in the technical literature, in applied work calibration has usually involved a simple model of the display considering only first-order variables. That is, the display is considered to have three independent color primaries so that one can predict the tristimulus values of any displayed light by knowing only the tristimulus values of the red, green, and blue primaries at their maxima and their display transfer functions. The tristimulus values of the primaries can be measured directly or calculated from the measured chromaticities and luminances. The display transfer functions are determined by measuring the luminance of each primary at a number of digital data levels.
Three approaches, in increasing order of expense and accuracy, are:
color calibration data are available one can use them to display the
desired physical colors. There are several options for using the data.
The most direct is to calculate the desired digital image data
in the application software, transforming as necessary from the coordinate
system used for the color design. If the application involves more
than one output medium (e.g., display and print) or input medium it
might be useful to use the operating system's color management software.
For example, one might feed the measurements to a program that generates
a custom ICC
display. Commercial software for that purpose is available from a
number of vendors.
Luminance Contrast in Color Graphics