However, searching prisms unsuccessfully at Edmund Scientific I did hit on the idea of trying to do it subtractively rather than additively. This way you only need two images.
You can easily get a plate beam splitter that's fairly large to split the incident beam. The transmitted beam will be reduced by slightly more than one f-stop, and the reflected beam will be approximately the same but oriented improperly because of the reflection. You'd need a front surface mirror to reverse the orientation of the reflected beam.
Several email exchanges with the Edmund Optical support team agreed that the beam splitter shouldn't affect the optical quality of the image produced by the lens in either the transmitted or reflected image. So long as the light paths are equal you'll get two images focused on the targets, each with slightly less than half the light of the original incident beam. (There is some loss in the beam splitter, but not a lot if it's professional quality optical gear.)
Plus you'd need the proper filters to make your negatives carry the right information.
And finally the reproduced color rendition isn't spectacular, just pretty good. This is why Technicolor's three color system blew away the two color systems that preceded it. Three color additive is capable of rendering true colors across the spectrum. Two color is a trick to get us to perceive missing colors that works because of how our eye/brain system is constructed.
Making a two color system at home isn't isn't trivial either, or I'd be all over it. But it isn't outside the realm of hobbyist possibilities either. And a plate beam splitter and front surface mirror is a lot less expensive than getting the tri-color prism.
Of course, you could always take three images with filters, but that's not what the thread is about.