Colour Correction

If you digitise your pictures yourself, from prints or negatives, you might have problems with getting the colours right. Below is a picture that has a nasty cast across it.

There are many ways of adjusting colour in an image, and if you find yourself doing a lot of colour correction you will probably find it useful to buy a book on the subject. I recommend Professional Photoshop: The Classic Guide to Color Correction (Dan Margulis). In the meantime this page describes a few simple steps that will help you remove colour casts caused when scanning images.

Colour correction generally involves identifying reference colours. Of course the most obvious reference colours are black and white. In an RGB (Red-Green-Blue) colour space, black generally means minimum of all three colours and white means maximum of all three. However when it comes to printing, most printers are not RGB, but CMYK (Cyan-Magenta-Yellow-blacK). Because of the different ways in which shades of colour are created on paper and on screen, there are colours that can be represented on screen than cannot be represented on paper, and vice-versa. Also pure black can be represented in many different ways within the CMYK colour space because black can be printed by either using black ink, or by combining equal quantities of the other inks, or by using a little of all the inks, including black.

Anyway, when you colour correct an image, the first thing to do is to identify:

  • the brightest white thing in the image, i.e. the part of the image that most closely approximates to pure white
  • the darkest black thing in the image, i.e. the part of the image that most closely approximates to pure black

In both cases bear in mind that light sources (such as the sun) and reflections (e.g. sun bouncing off a window) do not count. Neither do shadows. The reason is that although these may well prove to be the brightest and darkest parts of your image, they are not necessary pure white or pure black and cannot therefore be used as reference colours. Instead you need to identify part of the image which you want to appear to be pure white or pure black, such as the white T-shirt of a person featured in the picture, a white window frame, etc. It is of course possible to have an image in which there are no reference whites or blacks, but this is quite unusual. For those images you will have to look for a reference grey, or colour correct by other means. (I will not describe other technique for colour correction here - there are whole books on the subject as I mentioned above.)

In the following examples I am using Adobe PhotoShop. Similar controls exist in other applications. From the curves dialogue, note the pipette colour picker icons on the bottom right (see below). If you select the one coloured black and then click on part of the image, the curves are automatically adjusted so as to make the chosen colour black. The rest of the image's colours will shift accordingly. Likewise you can use the picker coloured white to select part of the image you want to be white. If you happen to know there is something in the image that is a pure grey (i.e. equal amounts of the colours) you can use the middle colour picker on it.

The task of identifying the brightest white part of an image can be made easier by adjusting the threshold of the image temporarily. For the above image, the threshold dialogue in Adobe PhotoShop looks like this:

Providing you select Preview in the threshold dialogue, the image above is shown like this:

Original image without adjusting threshold.


If I increase the threshold to close to the end of the displayed curve, the majority of the image goes black, revealing only the brightest parts of the image. By adjusting the threshold I can therefore see that the brightest white part of the image is the white of the window frames on the right side of the street.


By doing the same thing for black, by reducing the threshold, I can find the most black part of the image, which is the left side of the lamp post.


By carefully going back to the curves dialogue and selecting these two colours with the colour pickers for black and white, the image can be colour corrected as below)

Original Colour corrected