Comments on: Experiments in Ray-Tracing, Part 9: Interpolation http://www.hackification.com/2008/09/07/experiments-in-ray-tracing-part-9-interpolation/ Fri, 23 Jul 2010 14:53:52 +0000 hourly 1 http://wordpress.org/?v=3.0.1 By: kometbomb http://www.hackification.com/2008/09/07/experiments-in-ray-tracing-part-9-interpolation/comment-page-1/#comment-27 kometbomb Wed, 01 Oct 2008 18:57:42 +0000 http://www.hackification.com/?p=36#comment-27 Oh and I forgot to mention you can achieve the checker pattern also by interpolating the UV coords you have of the plane, that way you need to raytrace less pixels, the hardware/fast software mapping handles the rest (and again you can interpolate the normals etc.). Oh and I forgot to mention you can achieve the checker pattern also by interpolating the UV coords you have of the plane, that way you need to raytrace less pixels, the hardware/fast software mapping handles the rest (and again you can interpolate the normals etc.).

]]> By: kometbomb http://www.hackification.com/2008/09/07/experiments-in-ray-tracing-part-9-interpolation/comment-page-1/#comment-26 kometbomb Wed, 01 Oct 2008 18:55:39 +0000 http://www.hackification.com/?p=36#comment-26 You can regain that detail in the reflections if you simply keep track which object you can see reflected. Note, that it might be better to keep track of which object there is under a pixel, not what color values it has. I gained a lot of speed by doing a 16x16 initial check on a real-time raytracer I did, then recursing using smaller grids. The biggest problem was that if an object was smaller than the 16x16 and was aligned right between the initial pixels, it would not of course be rendered at all since none of the surrounding four pixels would know there is an object between them. A slightly better approach is to use non-square interpolation. Triangles work nicely. Also, since anti-aliasing was mentioned: the same technique can be used with a 2x2 grid to find the edges of objects that should then be anti-aliased (and the insides can be interpolated as described in the article), and the original method of comparing color values works even better in that. You can regain that detail in the reflections if you simply keep track which object you can see reflected. Note, that it might be better to keep track of which object there is under a pixel, not what color values it has.

I gained a lot of speed by doing a 16×16 initial check on a real-time raytracer I did, then recursing using smaller grids. The biggest problem was that if an object was smaller than the 16×16 and was aligned right between the initial pixels, it would not of course be rendered at all since none of the surrounding four pixels would know there is an object between them. A slightly better approach is to use non-square interpolation. Triangles work nicely.

Also, since anti-aliasing was mentioned: the same technique can be used with a 2×2 grid to find the edges of objects that should then be anti-aliased (and the insides can be interpolated as described in the article), and the original method of comparing color values works even better in that.

]]>