Wednesday 12 December 2012

Colour Separations

This is something to try whether you are using film or digital. The technique goes back to Maxwell so is pretty old, and at least in some circumstances is the only way to get a colour photograph from a monochrome camera. The Viking Mars landers used the same method to take colour pictures on Mars. In fact the system wasn't initially calibrated properly, so the colours returned in the first few pictures were wrong, showing Mars's pink sky as blue rather than pink.

Colours and filters

The technique is straightforward. With a camera set firmly on a tripod, take three or more identical monochrome pictures with different coloured filters, and then scan or otherwise combine them to make a coloured image.

The traditional filters to use are red, green and blue, and the traditional kind of filters are "gelatine" filters (so-called because they were once, but no-longer made from gelatine). I found a number of Kodak "Wratten" gelatine filters and made cardboard mounts for them to fit in a Cokin "P" holder.

Colour separation filters

For the record, the gels I used were: (Wratten numbers) 25 (red tricolour), 61 (deep green tricolour), 47B (deep blue tricolour), and I also tried 34 (violet). There are a number of possibilities, and the combinations I used are perhaps not optimal, but rather more limited by what gels I could source at a reasonable price. You can try anything, but strong blues are quite difficult to find. It might be worth trying a blue colour conversion filter such as 80A but may not be perfect.

Or go for something completely different, and get an unusual effect!

Trying it out

I chose a moderately colourful subject to try it out with.

With tungsten and DSLR's white balance

The technique is straightforward, and just as described above. Of course the camera must not move between exposures as you are swapping the filters over. This is a montage of my three coloured shots and (bottom right) the combined image.

Colour separation

I used the GIMP to combine the three images, by putting them in three separate layers and then altering the transparency of the layers. These layers were monochrome (desaturate if you didn't do so already) and then colourised to the obvious colours of red, green and blue. (For three layers, they contribute equally to the final image when transparency is 100%, 50% and 33%) In hindsight, this wasn't the best method to use as each layer then is 1/3 of its original intensity. It would have been better to simply add the pixel values in the layers but I didn't know (and still don't know) how to do this in the GIMP.

The final result (bottom right) is low in saturation, because of the problem just mentioned but IMHO has rather beautiful colours, despite the very poor lighting I was using for this quick test. Here it is when the saturation problem is resolved.

Combined image

Red, green and blue make white, right?

Finally, a little bit of science. As a teenager I learnt two things from my Physics teacher. (A) A colour of light is just a frequency of light and all colours are essentially equal - nature doesn't choose any one over any other - and white light is a mixture of all possible colours. (B) The three "primary" colours when mixed make white light.

This confused me greatly. I knew enough in the world of maths and electronics to know that component frequencies can be extracted from a mixture, so three specific frequencies (red, green and blue) cannot make the same result as a mixture of all frequencies. So how can red, green and blue make, and also not make, white? My instinct was to believe (A) and the maths, but we were also shown experiments that seemed to prove (B) was right.

The answer is not in the Physics classroom, but in the Biology classroom. And also, maybe, in the English classroom. For when some people say "the three primary colours when mixed make white light" they don't mean this at all, and should be more precise. They mean that when mixed they make something indistinguishable by the human eye from white light. And the eye has three colour sensors - red, green and blue. It is because of Biology, not Physics, that we can separate colours into three and then recombine them again. Aliens from another planet, or indeed other animals on this planet, would probably say (if they could speak!) that our colour images are at best very poor imitations of the real thing.


  1. Hi there... I stumbled across your blog today! I'm just starting out an adventure into camera modding, so your posts have been really interesting. Anyway, I just wanted to contribute: making "additive" images is quite easy in GIMP. If you have each color on a separate layer, simply go to the top of the layer window and use the drop-down to change the mode from "normal" to "addition". (do this for each layer) Hopefully that helps!

    I look forward to browsing your archives and seeing your future posts!

  2. After reading your blog, I wish I had not sold my Wratten filters. This is a great way to put them to new use.

    It is possible to get the red, green, and blue images to combine additively in Gimp. First make sure each image is saved in grayscale. Then create a new, empty image in the RGB format. In this newly created image, go to the channels tab and highlight exactly one of the channels, either red or green or blue. Make a copy of the corresponding grayscale image (Ctrl-C), and paste to the RGB image (Ctrl-V). This creates a floating layer. Once you have positioned the layer where it needs to be, select the Anchor option, and the floating layer goes away while the image is stored in the channel. Repeat this for the other two colors. When finished, you will have a full-color RGB image. You can then adjust brightness and contrast on each of the channels separately.

  3. Hello, I too stumbled on to your blog and read your experience with color separations. I became interested in this because of the Sergy Prokudin-Gorskii: Russian Empire photo archive at the Library of Congress. After piecing together several articles I figured out how to combine the RGB 'negatives' to form wonderfully luscious rich 'technicolor' photos. I wrote up step-by-step instructions on how to do it in GIMP on my blog. You might find it useful:
    Next I'm going to try making color separations myself. I acquired a Nikon N80/F80 35mm auto-advance camera. It was the last film Nikon made that had a cable shutter release. I'm working out how to make a 'universal' color filter adaptor so that color separations can be made quickly to minimize registration problems.