Jump to content
  • Advertisement
Sign in to follow this  
mv348

question on skinning with 'relative deformations'

This topic is 2253 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Okay. So I perfectly understand the usual skinning system. You have an initial bone matrix B[sub]i[/sub] which gives the initial bone-to-mesh transformation. Then you have a current bone-to-mesh matrix B[sub]f. [/sub]To compute the shift from initial(bind) pose to current pose, you are interested in the transformation T such that T*B[sub]i[/sub] = B[sub]f[/sub] so solving gives T = B[sub]f[/sub] B[sub]i[/sub] [sup](-1)[/sup] (Order of multiplication may need to be reversed depending on what system you're using, but same elements)

Okay thats fine. However! In the graphics package I am currently using, skin deformations also inherit from parent deformations. So the deformations from a bone are relative to other deformations. Call the initial/final parent deformations A[sub]i[/sub] and A[sub]f[/sub] respectively.

I am extremely puzzled because the transformation the example code uses is: A[sub]i[/sub]A[sub]f[/sub][sup](-1)[/sup]B[sub]f [/sub]B[sub]i[/sub] [sup](-1)[/sup][sup] [/sup] Notice it uses the inverse of the final matrix. I would have expected the inverse of the initial to come first.

I guess its worth mentioning that the complete transformation is carrried out in a particular node's space with transformation M so the total transformation formula is:
M[sup](-1)[/sup]A[sub]i[/sub]A[sub]f[/sub][sup](-1)[/sup]B[sub]f[/sub]B[sub]i[/sub][sup](-1)[/sup]M

But that matrix M is not time dependant so I don't think this matters much.

Its also worth mentioning that the documentation did not give a very clear explanation on exactly how deformations are inherrited. So if anyone has a good hypothesis on that, for which the given formula makes sense, I'd love to hear it!

(edit) So I came up with a HAZY idea of what might be happening. Assume there is no bone shift (B[sub]f[/sub] = B[sub]i[/sub]) so the transformation is just A[sub]i[/sub]A[sub]f[/sub][sup](-1)[/sup]and consider the following diagram:

associate_model.png

On one hand, this can be interpretted as the coordinate system 'A' shifting to the right. But if the coordinate system is viewed as the point of reference, then the mesh is actually shifting to the left. This would explain the reverse order shift from A[sub]f [/sub]to A[sub]i[/sub]

The exact story is still very blurry to me though. Anyone that can offer some insight here would be very much appreciated! Edited by mv348

Share this post


Link to post
Share on other sites
Advertisement
Sign in to follow this  

  • Advertisement
×

Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

We are the game development community.

Whether you are an indie, hobbyist, AAA developer, or just trying to learn, GameDev.net is the place for you to learn, share, and connect with the games industry. Learn more About Us or sign up!

Sign me up!