Long ago I wrote tools for that purpose. Not that difficult to do.
I just started with the standard color blindness tests, measuring the patches. Converted to Lab. Those patches implied a 2-manifold in Lab that the protan (or deutan) can "see". Any other color gets mapped onto that 2-manifold. So I built a 3-d lookup table that did the mapping (again, all as Lab->Lab mappings.)
I used this scheme to build protan and deutan specific mappings, I think I did it for yellow-blue color blindness too, but it has been 25+ years since, so I no longer have any of these mappings around. (For those of you who remember the Kodak ColorSense package, these were all in the form of effects transforms.) I don't recall if the manifold was close enough to a planar subspace or not. But I do recall that subspace was pretty simple.
It did produce very pretty pictures that emulated what a person with the given variation of color blindness would see.
