Jump to content
  • Advertisement
Sign in to follow this  
timw

smoothing normals part II c-1 continunity

This topic is 4829 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

I thought of that, using a bezier space, much like what is used in original free form deformation implimentation. in this way, it would be continuous for all points in space, and thus it has to be continuous for points on the surface. what do you think? maybe I can interpolate using radial basis functions, does anyone know if radial basis functions are c-1 continuous? toji that was what I intened to do originally, but the problem is when I subdivide the mesh I have to do more and more physics calculations for my cloth simulation and that quickly becomes VERY VERY slow, considering I'm using a simple first order euler integration technique. I know I said I wanted off line algorithm, but I still want the physics portion to be fast so I can visualize my results quickly, I'm gonna render using a form of path tracing so the rendering is offline, but I want to be able to see the cloth quickly to find pleasing draping positions for which to render offline. thanks for the advice I'll check out that article Tim [Edited by - timw on August 24, 2005 5:03:18 PM]

Share this post


Link to post
Share on other sites
Advertisement
As far as I know when using the fixed function pipeline the only way to really smooth things out is to subdivide the mesh more. I suppose that if you were using shaders you could recalculate the normals per-fragment using a simple of bezier curve algorithm. That could give you quite interesting results where the surface would look very nice internally even if it's a little rough around the edges. (Actually, I'm tempted to try that now!)

Still, the bottom line is that if you're using this for offline rendering the traditional solution is subdivide subdivide subdivide!

Share this post


Link to post
Share on other sites
You're going to have to fit the data to some sort of spline or bezier surface and interpolate your normals using that surface. Take a look at this article at gamasutra about N-Patches.

http://www.gamasutra.com/features/20020715/mollerhaines_02.htm

Share this post


Link to post
Share on other sites
I thought of that, using a bezier space, much like what is used in original free form deformation implimentation. in this way, it would be continuous for all points in space, and thus it has to be continuous for points on the surface. what do you think? maybe I can interpolate using radial basis functions, does anyone know if radial basis functions are c-1 continuous?

toji that was what I intened to do originally, but the problem is when I subdivide the mesh I have to do more and more physics calculations for my cloth simulation and that quickly becomes VERY VERY slow, considering I'm using a simple first order euler integration technique. I know I said I wanted off line algorithm, but I still want the physics portion to be fast so I can visualize my results quickly, I'm gonna render using a form of path tracing so the rendering is offline, but I want to be able to see the cloth quickly to find pleasing draping positions for which to render offline.

thanks for the advice I'll check out that article

Tim

Share this post


Link to post
Share on other sites
timw,
I did a similar solution once with a cloth sim. I used the grid of simulated cloth points as a mesh of control points for a cubic surface (I used Catmull-Rom for simplicity's sake), and calculated the rendered cubic surface after doing the simulation step.

So simulate N points, then subdivide to 4N points (for instance) and render those.

I was doing real-time though, and decided it was fine to just go with raw vertex normals. If I were doing offline, I would use a nicer surface equation than Catmull-Rom, probably some subdivision surface.

Share this post


Link to post
Share on other sites
interesting all this time I didn't even think of using a different surface representation, lol I was trying to do some normal trick. sounds like a good idea, thanks for the response.

Tim

Share this post


Link to post
Share on other sites
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!