6 replies to this topic

[rouncer]

Whats the quickest way to go hipoly->lowpoly. the way ive done it up till now is pretty wasteful, it takes continual reiteration of the mesh slowly sucking out triangles each pass, whats the professional way to do this?

[rouncer]

im reading this -> http://en.wikipedia....gressive_meshes
but its not helping. : P

[rouncer]

this is better, but its looking like there is no quick way to do this unfortunately.
http://graphics.stan...plification.pdf

[rouncer]

I was thinking, would there be any way to do this on the video card?? like with the geometry shader?

[rouncer]

ah found it! http://gfx.cs.prince...plification.pdf

[rouncer]

any help?

[rouncer]

I looked at their implementation, and it would destroy long thing areas of the model, id like to preserve these.

im coming up with my own algorythm-> which is ->


STEP 1.

make texture or vertex stream, either way would work... of index list of adjacent triangles for each vert your going to test to collapse on cpu. you dont collapse two vertices adjacent to each other in a single pass.

STEP 2.

on gpu (geometry shader with stream out) -> collapse verts accessing index lists and reoutput collapsed triangle stream. the triangle streamout is indices, not vertices.


you need a texture with the vertex positions on it, to keep it index based. (so the model doesnt separate apart from itself)


only problem is, its half and half cpu gpu, and id like the whole thing gpu, any one could help me out getting step 1 to work on the gpu?
there could be some max amount of indices per collapsable vert, that could be ok...



Im thinking maybe you could write some collapsive restorative poly modeller with it, imagine tesselating data then pulling points, then re edge collapsing it after the edit. then collapsing the next tesselation, etc,etc... that could be insanely powerful! possibly 3d accellerating mesh simplification could be pretty cool.
Just a thought tho.