Change the way in which multi-color images are handled

[cjfuller]

Currently, images with multiple channels are split and only one channel is processed (though all may be quantified). Add the ability to process all channels, or to keep them together so that methods can do special processing drawing from all the channels.

One possibility would be to have the ImageSet class create a single image containing all the channels and pass that to methods, though this would break the interface to current methods and require rewriting.