The simplest effect which we might require is the following: since some colors (of source gamut) are reproducible (in the destination gamut) and others are not, we want to reproduce the former exactly as they are, while the latter are replaced by the nearest reproduceable color at the outer limit of the destination gamut (in short, out of gamut colors get clipped). This means that colors that were different in the original image can after the conversion be the same.

For some types of image, this is an adequate rendering. For example, when we know for sure that all the colors in an image are printable, which is often the case. The typical example of this is the company logo or brand name. Imagine the Coca Cola red or the IBM blue; whoever designed these logos no doubt thought about the printability problem and deliberately chose colors which are always printable. The Coca Cola logo we see on the monitor will undoubtedly have printable colors. Only rarely might it have one or two pixels with an unprintable color.

It is therefore reasonable in this case to print the printable colors (probably the great majority) exactly as they are, and print approximations to the rest (in an operation known as clipping). This type of rendering is called colorimetric.

Click here for a QuickTime movie showing how colorimetric rendering works. To see this you need the QuickTime plug-in installed on your browser: if you don’t have it, click here.

In practice, there are two forms of colorimetric rendering. The form described above is the absolute colorimetric rendering.

The absolute colorimetric rendering does not expand or compress the whole gamut: each color is transformed into itself, if it exists in the destination gamut. Otherwise, it is transformed into the closest color at the gamut boundary.

Note that the original white may or may not exist in the destination gamut: in the first case it remains unchanged, in the second case it is changed in a similar color. The absolute colorimetric is often used in print proof and where possible does not modify the brightness.

The relative colorimetric rendering requires instead an exact color match in everything but brightness, which may be modified so that all the brightness levels are within the range of brightness of the destination gamut in use.

With relative colorimetric rendering, the source white (e.g. 255R, 255G, 255B) is converted into the destination white (e.g. 0C, 0M, 0Y, 0K) (this is called white point compensation); all other colors are shifted accordingly. The resulting image may be lighter or darker than the original, but the white areas will coincide.

It is up to the user to decide whether absolute or relative colorimetric rendering is more suitable for the image he/she is working on.

If the destination gamut is wider than the source gamut, absolute colorimetric rendering is more suitable because the origin white is included in the destination’s range of colors.

If the destination gamut is narrower than the source gamut, it is usually best to choose relative colorimetric. If the two whites are not the same white (the source's one, perhaps monitor, is brighter than the destination's one, perhaps printer) the source's white is made to correspond to the white of the destination, which is normally the most sensible solution. Absolute rendering would produce a printed white which is an approximation to the white on the monitor.

In some applications the relative colorimetric rendering is referred to as "graphics" or "logo color", while the absolute colorimetric rendering is referred to simply as "colorimetric" or "proofing".

For some situation, neither absolute nor relative colorimetric rendering is suitable. This is the case when the destination gamut is narrower that that of the source (for example, when the source is a monitor and the destination is a printer) and the image is a photographic or "realistic" one.

In this case the gamut is to be compressed, but the colors must keep their relative chromatic positions; it is not acceptable that some of the colors might be exactly reproduced whilst others are only approximate. All the colors, even those which could be reproduced as they are, are to be altered (typical this rendering desaturates all colors), in such a way to maintain their overall relashionships, and the eye will be able to compensate for the difference between (for example) the image on the monitor and the printed image.

Click here for a QuickTime movie showing how perceptual rendering works.

This rendering always compresses the complete source gamut (not only the part which is not within the destination gamut) but preserves the relationship between the colors, and is known as perceptual or sometimes as "image", "photographic" or "photometric".

There is one last special case and it is, paradoxically, the case in which the precise matching of colors is of little importance. The typical example is that of statistical graphs, where it is more important that the colors are bright and saturated, than that they are exactly the same as the original.

Saturation rendering, sometime also referred to as "graphics", requires, as the name suggests, that the saturation of the colors be preserved in the transformation from gamut to gamut, perhaps at the cost of brightness and hue.

Click here for a QuickTime movie showing how saturation rendering intent works.

With this type of rendering, the original colors are modified in order to exactly fill the destination gamut. This means that some areas are compressed and others are expanded. This is the only rendering which can expand a limited gamut to something wider.