Produces four .png files with transparency, one for each of the colour separations cyan, magenta, yellow and black. For each pixel, the more saturated a separated colour is, the more opaque it is on the output file, this allows overlay in something like Photoshop or Illustrator which magically gets close to the original colour if all four are overlaid, I have found best in the stacking order from top to bottom of YMCK, I also reduce the opacity in Illustrator for each of the separations, otherwise the magenta and cyan in particular come out too overpowering.

The reason I wrote this is to experiment with overlaying print colour separations for display on screen. I could not find any other software that does this, at least not for free.

The way in which I separate the colours is not by all-purpose formula or lookup table. Instead, for each of the CMY colour separations one of two options can be selected, the first – absolute values – takes the RGB values for the appropriate pair, adds them together and gets the average, i.e. divides by two, then subtracts the value of the third of the triad (e.g. if we are doing cyan we get the average of blue and green and subtract red). If this produces a negative number then we make it zero. Black is simply the average of the three values in the triad, inverted by subtracting from 255 as with black we want a transparency not an opacity. The other option, by brightness, is similar except that instead of taking the RGB value, we use the brightness level for each element of the RGB triad, and in this case instead of the average we use the cumulative value to a maximum of 255 because brightness is cumulative, subtracting the third element of the triad’s brightness similarly as for the absolute values formula. Depending on which brightness calculation formula is selected, this seems to work better than using absolute RGB values – better in the sense that it produces a more convincing result when the separations are overlaid.

The CMYK values then aren’t therefore, as would normally be the way, determined by printer profile, because in this case there’s no printer involved, it’s simulation on screen, so a pixel-based formula is more relevant.

You can see some examples of output from this software on my Flickr photostream, e.g. at Colour Constancy Cyan.