The thing that strikes me as problematic for 3D compared to 2D is that in 2D the renderable representation directly embodies the edges that you wish to antialias. For 3D, this is not the case: the outline of a 3D object varies depending on its position with respect to the viewer. The so-called "silhouette edges" need to be recomputed each frame.
Silhouette edges are often used in artistic rendering styles (such as "toon shading"), but typically not for realistic-as-possible scenes. As you say, the standard practice is to use whole-scene super-sampling (actually multi-sampling... an optimization where only the depth and stencil buffers are super-sampled but the more expensive fragment shader is run only once per pixel). The closest thing to an analytical antialiasing approach for 3D (that I'm aware of) is the recent research on "morphological antialiasing". I don't believe that I've yet chimed in to congratulate Juan on his work. Very cool stuff! I'm looking forward to getting my hands on it. Cheers, Josh On Jun 7, 2010, at 5:52 AM, Juan Vuletich wrote: > Wesley Smith wrote: >> Is it possible for your system to treat 3D graphics? Would you >> project to 2D and then pipe it through Morphic? How would this work? >> wes >> > > Hi Wes, > > I've never worked with 3D. But given that what is usually done (I believe, > I'm not an expert) is whole scene super sampling, the quality / cost ratio > could be improved with my work. Of course, I'd need to run it in the GPU, to > beat OpenGL. I mean, there is a lot of work to get that done. > > Cheers, > Juan Vuletich > > _______________________________________________ > fonc mailing list > [email protected] > http://vpri.org/mailman/listinfo/fonc _______________________________________________ fonc mailing list [email protected] http://vpri.org/mailman/listinfo/fonc
