Sadly, not much information can be found about this important subject. It truly deserves a long discussion but – until somone writes up something longer – you will have to settle for this blog post!
Color-calibration is important to see accurate image colors and to best visualise photographs. Often, we get comments from people who notice that their images have different colors on different monitors. This happens because monitors are not all capable of showing the same colors and because not all monitors show the same color for the same input – even for the colors which they can display. The former problem is determined by the color gamut of a display, while the latter is controlled by color-calibration, or lack thereof.
The simple solution is to tell people to calibrate their monitors. So, a lot of users search the net and find ways to calibrate their video cards! Unfortuntaly, these are not equivalent, although most online articles refer to both as Display Calibration. The difference is subtle but quite important for users of LCD displays. For those lucky enough to still use a high-quality CRT, it is much less important.
Calibration works by translating the color-value of each image-pixel into the color-value used to display it on the monitor so that it appears as its intended color or so. Given an sRGB image and a perfect sRGB display, the translation should not change any values. However, since displays are generally not perfect, some translation is often necesary. Image pixels are represented as tripplets of numbers, one for each of the red (R), green (G) and blue (B) additive color primaries. In most cases, these numbers are 8-bit values running from 0 to 255. All but two current LCD displays only accept 8-bit values as well. This is a limitation of the DVI-D connection which carries in its protocol at most 8-bits per color-component. Analog signals used by CRT monitors do not have this limitation and neither does the new DisplayPort connector, which supports up to 16-bits-per-component, but only if the display can accept it.
The difference between calibrating a display and a video card comes from these numbers. Since almost every LCD on the planet can only accept 8 bits-per-component, calibration of the video card can only send 8-bit values to the display. This gives rise to precision problems since a translation from 8-bits-per-component to 8-bits-per-component has to create gaps and overlaps in its output. For example, say the red component is 1/4 too bright, red image values 7 and 8 will be both transalted to 6 for the display. Such precision problems will appear as banding on the display. The opposite – posterization – is also possible when the translation does not use all possible display values.
Display calibration avoids this problem to a large extent by doing higher-precision translation after the signal is sent to the display. In this case, image pixel color-values are sent unchanged from the computer to the display. All 8-bits-per-components are used, so banding and posterization are not introduced in the signal. Instead, the display translates from the 8-bits-per-component signal to a 10, 12 or 14-bit precision using an internal calibration table. To set this table one must calibrate the display and not the video-card. This requires a calibration device which plugs into the monitor directly and usually model-specific software to control it.
The catch is that not all monitors are capable of doing this. It used to be the case that only expensive high-end monitors had this ability but this is no longer the case. The cheapest calibratable LCD available now retails for $499 USD or $585 CDN. That would be the NEC Multisync P221W which has 10-bit internal tables.