User:Sonny Day

1. In theory - as a more or less analogue system - the human eye can distinguish an infinite number of shades within the overall range of perceivable colours. This 'overall range of perceivable colours' is effectively the 'colour-space' of the human visual system.

In very simplified terms visible light comes as a continously varying range of 'colours' ( cf. rainbows/prism spectrums). The eye however only has colour receptors/sensors for the three primary colours - red, green and blue (RGB). The eye encodes each colour as a unique combination of output levels from these sensors which it sends to the brain, which in turn perceives that colour. Mixing all three in 'equal proportions' gives white light. The eye/brain cannot distinguish between the 'colour' produced by simply mixing the appropriate proportions of just the three primaries and that due to the whole range of natural light. This is central to the long-established RGB colour model . It explains how TV screens/ computer displays etc. only need to generate three different colours and cameras equally only need three different colour sensors. As the intensity of each primary colour can be varied continuously, in theory at least an unlimited number of shades can be produced. See colour vision.

2. A digital recording system records an image as a finite array of pixels each of which has a specified colour. The colour is specified by recording the intensity levels - detected, for instance, by a camera sensor - of each of the three primary colours (RGB). However in a digital system, only a finite number of levels can be recorded (without using infinite data) which means only a finite number of different colours can be recorded.

A typical recording might employ '8-bit colour depth'. Only numbers between 0 and 255 can be written using 8-bit, so only 256 levels can be specified/recorded for each of the three primaries. So RGB [0,0,0] = black. RGB [255,255,255] = white . RGB [255,0,0] = saturated red etc.. 256 levels for each of three primaries gives 256 X 256 X 256 = 16,777,000 combinations/different colours. (often rounded down as 16 million colours). When light falls on a sensor its colour is recorded not as its true colour but as the closest to it of the 16 million possible 8-bit colours. Whilst each primary is represented by an 8-bit number, each pixel in RGB requires three of these numbers and is therefore confusingly called 24-bit

3. If you distribute 16 million colours evenly across the whole range of human vision, the differences between each discrete shade would be quite noticeable.

It is essential to have only small differences between shades in order to reproduce subtle details and provide a smooth transition between shades. This would require far more colours, requiring a greater bit depth - in other words much more data. (In fact 16-bit is frequently used in high quality work. This has 65,536 levels per primary giving billions of shades and confusingly with three 16-bit numbers to specify the colour of each pixel is often called 48-bit.)

4. However, many of the colours which would be specified in attempting complete coverage would be very dark or extreme shades. It turns out that we can still have acceptable colour rendition for many purposes if these shades are simply ignored.

This is fortunate as in reality an RGB system cannot actually reproduce all the visible colours - only those in fact which lie within the range de-lineated by the three chosen primaries. If you mix equal parts of red and blue you obtain a shade of magenta determined by the particular red and blue being used. You cannot make a more extreme shade of magenta by adding green, only by changing the actual colour being used for the red or blue primary.

By distributing the 16 million colours (of 8-bit) across less than the full range of human vision - specifically excluding the very dark and extreme shades referred to above - it is possible to make the difference between adjacent colours smaller - thereby achieving sufficient shade differentiation where it matters without having to increase the amount of data required.

5. As an approximate guide sRGB covers 30% of the visual range whereas aRGB covers about 50%. (Depending somewhat on luminance levels and also whether one allows for certain perceptual factors.)

As both have the same number of colours to distribute, the sRGB colour-space has better shade differentiation within its range whilst the aRGB colour-space covers some extra colors - more specifically, extended green range at low luminance levels and intense cyan/green, magenta, orange and yellows at high luminance.

These ranges are different because each colour-space uses different primaries - those of sRGB being closer together than those of aRGB. An important consequence of this is that the actual colour corresponding to for example (100,150,255) in sRGB is not the same as (100,150,255) would be in aRGB.

6. The loss of shade differentiation using aRGB is potentially significant at 8-bit but is not at 16-bit with its much greater number of levels and therefore colours to distribute.

7. At its simplest, the colour space of a device refers to the range of colours that it can either record or reproduce.

8. Displays and printers have their own colour-spaces. That is any given screen or printer can only reproduce its own individual range of colours and levels of shade differentiation.

9.The colour-space of your recorded image, your monitor and your printer may all be different.

This means that a given 8-bit RGB value corresponds to a different colour in each! In order to get the best colour reproduction, it is necessary to change the RGB values in the recording file to the one's which give the closest colour match in the monitor's or printer's colour-space. A very common example of this depends on the fact that most displays approximate to sRGB colour-space, so by converting your image file to sRGB, it will display acceptably and then one can tell the printer that the incoming file is in sRGB, which it knows how to interpret. Colour profiles can be assigned to files which may enable these changes to happen automatically - a process known as colour management.

10. Good quality printers have a colour-space that exceeds aRGB and therefore will print the extra colour range if it exists.

11. Lesser quality printers and some commercial printers have a colour-space that is smaller than aRGB and so will not be able to reproduce these extra colours.

12. The majority of computer screens have a colour-space more or less the same as sRGB and cannot reproduce the extra possible colours of aRGB.

13. UNLESS an image contains deep dark greens and intense yellow / orange / magenta which are important to the final image AND the image is to be printed on a high quality machine which can utilise aRGB then there seems little point in using aRGB and many reasons to prefer sRGB. In any case if ones shoots in Raw and retains the file, then the data required for aRGB will not have been lost.

14. In order to display on a sRGB monitor with any accuracy, an aRGB file must be converted to sRGB (essentially this means re-allocating the colour of each pixel in aRGB colorspace to its nearest equivalent colour in sRGB).

15. Having converted a file from aRGB to sRGB, the wider gamut colour information is lost and cannot be retreived by reconverting to aRGB. Equally the greater shade differentiation of sRGB will be lost on converting to aRGB.

16. Assigning a profile is not the same as converting.

Assigning merely tells a printer or display to treat the data of one colour-space as though it were from another. Most commonly, this occurs as assigning an sRGB profile to an aRGB colour-space image, with the result that the colours are incorrectly reproduced (typically washed-out).

17. Commercial photolabs very commonly work in sRGB. It is not unheard of for commercial printers to assign sRGB to aRGB images rather than converting them - with resultant colour errors.

18. If your browser is not colour managed (at present only Safari and FF3 are) it will assume that any image file is in sRGB colourspace. Since most computer screens are very close to sRGB in colourspace, then an sRGB colourspace recording (file) will reproduce reasonably accurately. (Assuming of course that the screen has at least got the colour, brightness and contrast controls adjusted correctly, if not a proper calibration.)

19. If you shoot in aRGB (or in Raw and then convert to aRGB), your monitor will still show the sRGB version unless you have a wide gamut screen, so you will not be able to see any difference until you print. Generally it will an unexpected boost to the image, not damaging.

20. Unless you have a definite reason, you should always use sRGB on the web. Shooting/saving in Raw covers a wider colour-space with better shade differentiation than either sRGB or aRGB. You have to have a good reason not to be using Raw.This is a function of the colour management profile.