Here is the result of an elbow test i'm making so far. I think it will need some paint weight refinement.
Something that is starting to work for me when creating cartoony deformations is to start really rough (flooded skin weights). Then i try to fix the most evident issues that a skin weighting would probably not be able to correct. I'm primarily using set driven keys where a primary skin joints rotation is telling a possible helper joints translations where to go. One advantage of doing this rough skinning work with all joints to start with allows me to make topology changes and still not lose previous progress.
Here are some images i made of some of the process i'm still working on refining.
basically i started really rough just trying to see roughly if i can prevent extreme collapsing with a broad set driven key pass (top left). Then i noticed topology in elbow could be improved so i mark the area add the new topology (bottom right). (note: should be able to preserve most of old weighting with Maya's copy weights tool, its helpful to do a quick glance at weights to make sure all were copied over well).
After rough skinning of essential areas complete i'll go in and try to clean up shapes (possible weight refinement and possibly if needed add corrective shapes). Im still trying to work on a deformations for face workflow.
Rich Diamant (online tutorials on deformations and anatomy)
And here are some more of my ramblings about the question Can we reuse Blendshape after changing topology? These ideas are probably very rough and need alot of development to be used in practice.:
Can we reuse Blendshape after changing topology?
(limitation would lose any smoothing modeling done on new topo mesh, it might be able to use the new topology though)
-- use ZBrush project old default onto new topo default mesh (hopefully doesnt change the new topology),
make old default wrap and move the projected new topo mesh.
make blendshape into old default mesh.
when blendshape on smile activated it should get old mesh to move to smile >> this gets the projected new topo mesh to move to smile
copy out the smile on new topo mesh (should work even if brow subtly altered)
-- is there a way of doing a triangulation of two meshes where both triangulations have idential vertex ordering ? Is this a different harder problem then our intial question
-- formulating blendshape reuse as a statistics prediction problem
Predict position for every vertex of previously made blendshape ?
--- some vertices we know prediction 100pct because topology didnt change
Predict position for every vertex of previously made base mesh?
-- might be simpler problem
But how can it be a statistical problem if just have one sample?
We have essentially 2 meshes to work with.
Somewhat related: How to reuse corrective shape on new topology mesh for correctives?
suppose ive added corrective shapes to a mesh. but then i change a new mesh and put all joints and weights on new mesh. Can we get all the correctives automatically working on new mesh?