PDA

View Full Version : Issues with stitching



Xarragon
07-12-2004, 02:17 PM
Hello, I'm currently implementing some basic skeletal animation with simplified skinning ("stitching"). I have a continous mesh, with vertices attached to individual bones. Those triangles with vertices attached to different bones will be stretched during animation, avoiding "scars" in the mesh. Each vertex can only be attached to a single bone, hence the term "stitching".

The issue I'm concerned with for the moment is the normals for those faces that are stretched. I have looked around a bit for a solution to this issue, but havn't really found anything as of yet. The closest thing I have found is this demo with source: http://home.planet.nl/~monstrous/dl/skelanim.zip

In this case, the author simply seems to use the original normals for all faces, but I could be wrong on this.

Two obvious solutions are aviable:
1) Recalculate normals on update. This however could be very expensive if we have many faces and want to average the normals.

2) Ignore the issue, and hope that it looks good most of the time. This could work if the deformation isn't too extreme.

I'm really asking because I thought there might be some "magical" solution, such as a result of the two matrices and points somehow could give a good approximation for the normal. A real shot in the dark, heh...

Anyway, thanks for any advice you could give me on this.
Martin "Xarragon" Persson, Sweden

Xarragon
07-13-2004, 05:11 AM
I just realised... I'm a big fool, right... I mean, I assign normals on a per-vertex basis, and each vertex is tied to a bone (read: transformation matrix), so it isn't an issue at all. OpenGL will do the acctual work for me during rendering.

(Note to self: Grow some sort of basic brain.)

You may make fun of me at your discretion. :D