Since the main goal of most rendering processes is to display the image for human observation, we should examine some display media characteristics in order to use this media properly. Rendered images can have a dynamic range of several thousands, and even more. As stated in the previous chapter our visual system operates in an impressive dynamic range. Unfortunately, display media dynamic ranges are quite small. For example, a CRT monitor, which is the most widely used display media in computer graphics has a dynamic range of up to 100! Obviously, huge dynamic range raw images should be mapped somehow to the relatively small dynamic range of display devices. An ideal display media would have a dynamic range that equals that of human vision capabilities, and would have the possibility of displaying luminances as low as the threshold of human vision, and as high as the maximum still perceivable luminance. It should be capable of reproducing visible colors, as well. Although such devices have long been in existance in acoustics, they will not be available in the video media for a long time.
There are two kinds of display media, light-emitting like CRT or, in a way, projected slides, and light-propagating, like photos or prints, which do not emit light themselves. Light propagating media is suitable for displaying solid colors by means of an external light source, while the other group has gamuts exceeding the solid colors and has the capability to display more saturated colors. E.g. the saturation of CRT blue can never be achieved with photo paper.
There are three major problems concerning display devices. The first is the display devices' non-linear response, the second is the limited dynamic range and the third is the limited color gamut.
Practically all display media have nonlinear characteristics. Fortunately, that is not such a big problem, as long as the user is aware of it. If the characteristic of the device is known, some correction can be done, and the device will act as a linear device. The situation is more complex for display chains. Let us take a chain of CRT negative color film photo paper, or a chain film-writer color slide photo paper. It means, an image is displayed on the monitor, then photographed, and finally a print is made from color negative film (this is valid for the first chain). For such chains the last link is important, and inputs to the first link should provide the desired results at the chain's end. In other words, if the photo is made from a slide, which is made by a film recorder and the input to the film recorder is a TIFF file, it is not important how this TIFF file looks on the CRT, or when the slide is projected. All that matters is that the final print looks satisfactory. Of course the final print depends on many variables. They include the type of film used, the quality and temperature of the developing chemicals (in the above case there are two developments), the length of time the film is in the chemicals, the type of photo paper used etc. There are also some influences from one color channel to others in all color media, some additivity failures [ToHe89] and so on. It is obvious that it would be quite difficult, if not almost impossible for the common user, to take into account all the above mentioned difficulties. What the common user should do, is to check the linearity of the device (chain) and the available device contrast.
The second, and more complex problem is the limited device contrast. As mentioned earlier, huge contrast raw images should be mapped to relatively small contrast display devices. This is, actually the most challenging part of most tone mapping techniques. We have measured CIE Y values of the photography made from the slide from the film recorder. Figure 3.1 shows the results. Input values to the film recorder are on the x axis, and the CIE Y values of the last chain link - photo, are on the y axis. Achieved contrast was 74.7/4.0=18.675.
Figure 3.1: Measured characteristic of a photograph print made from slide
Typical device contrasts are given in table 3.1, the ideal values and values that we have measured on common available devices are given in the table. Note quite a big difference in photo contrast. We have measured the contrast of photos that are automatically processed by Kodak. Some professional laboratories in Vienna are offering better service, but at, approximately, six times the cost of the usual Kodak service. We assume that the contrast in this case would be greater, and closer to the theoretical maximum.
|Display media||Typical contrast||Measured contrast|
|Newsprint printed in B/W||10|
The third above mentioned problem is the display devices' limited color gamut. There is a lot of research done in gamut mapping. Throughout this work, we will apply simple color components clipping, in case they exceed the device gamut. This approach can lead to hue changes in some cases, but more advanced gamut mapping techniques are out of the scope of this work, and can be found in [GeAl89], [HoBe93], [WoAB94].
The most widely used media for computer graphics are the CRT monitor and color printer, for sure. Figure 3.2 shows the color gamut of a typical monitor, and a printer. The gamut of a printer using highly saturated inks is shown as well. It is obvious from this figure why is it impossible to reproduce monitor images on the color printer perfectly.
We find the slides and CRT the most interesting media, therefore they will be explained in a simplified, yet for us sufficient way next.
Figure 3.2: Color gamut of printer and CRT