Categorising images by their dominant colours

I'm looking to categorise some photographs by their dominant colours. I will have a folder containing a number of images. I then want to have somewhere to define a list of colours to look for, and then have the code go through all the photos, see what the dominant colours are, and then output a list of the photos and the name of the colour groupings that most closely match that photo to <STDOUT>. A colour grouping is one or more RGB values. So a colour grouping may simply be a defined green, in which case photos that are mostly green should be tagged with that, or a colour grouping may be two colours, red and green, and photos which are mostly red and green should be tagged as that. If a photo matches more than one colour grouping (it's red and green, so it matches a red, a green and a red and green grouping) then that's fine. Just output all that. It may be good to have some thresholds too, to output a numerical value of how close it is, in case it matches two colour groupings, then I can decide which one to use by whichever is the closest. The app needs to run on a Linux server from the command line. I'd like if it was written in Perl at best (as that's my preferred language), but really any language is fine as long as it's fast. And if it's going to use some external software to help (such as ImageMagick etc) then that should be easily installable on Debian. I'm happy to follow any instructions to get something going on my own servers. It'll need a config file, so, in hash/dictionary terms something like this I guess, to give you an idea... [ { colours => [ [ 255,0,0 ] ], name => "red" }, { colours => [ [ 255,255,0 ]], name =>"yellow" }, { colours => [ [ 255, 0, 0 ], [ 0, 0, 255 ] ], name => "redandgreen" }, ] So really I just want to be able to cd into the folder of photos, run the command, and have it spit out it's output.