Specular convex reflector and surface normals?

Hi,

one question regarding "rounded" surface normals on specular surfaces. I have a dome with a highly specular (consider 99% specular) surface. The dome is an obj-mesh, with interpolated vertex normals ("smooth") and imported using obj2mesh.

The rounded vertexes render a smooth dome when viewed from below, as expected. But as this dome reflects light onto the floor, it is some kind of a big convex reflector. I wonder if for light reflected from this reflector to the ground, the normal interpolation is applied, too. The background for my question are big bright splotches on the floor. I also wonder if this geometry is something not really to be modeled in classic radiance, as the old questions of convex reflectors and backward raytracing might occur. Is this even a case for photon mapping? I have never been in modeling a convex reflector, for architecture this is not a typical example in our days, so I am curious about what you propose...

CU Lars.

Hi Lars,

If you use the "mirror" material type for the ceiling, it will cast appropriate source reflections, but without the surface normal interpolation. Also, it will incur a lot of calculation overhead, proportional to the number of triangles in your ceiling. I don't think I'd recommend it for that reason.

Surface normals are interpolated for other parts of the calculation, such as the diffuse indirect, which in your case is not so diffuse. This is no doubt the source of your splotches, and the photon map addition may well take care of the problem. You could try it with the "mirror" type first, just to see how long it takes and what it looks like.

-Greg

ยทยทยท

From: "Lars O. Grobe" <[email protected]>
Date: May 14, 2007 1:10:30 AM PDT

Hi,

one question regarding "rounded" surface normals on specular surfaces. I have a dome with a highly specular (consider 99% specular) surface. The dome is an obj-mesh, with interpolated vertex normals ("smooth") and imported using obj2mesh.

The rounded vertexes render a smooth dome when viewed from below, as expected. But as this dome reflects light onto the floor, it is some kind of a big convex reflector. I wonder if for light reflected from this reflector to the ground, the normal interpolation is applied, too. The background for my question are big bright splotches on the floor. I also wonder if this geometry is something not really to be modeled in classic radiance, as the old questions of convex reflectors and backward raytracing might occur. Is this even a case for photon mapping? I have never been in modeling a convex reflector, for architecture this is not a typical example in our days, so I am curious about what you propose...

CU Lars.

If you use the "mirror" material type for the ceiling, it will cast appropriate source reflections, but without the surface normal interpolation. Also, it will incur a lot of calculation overhead, proportional to the number of triangles in your ceiling. I don't think I'd recommend it for that reason.

No, the geometry is awfully complex, consisting of a number of domes and semidomes which are connected by rounded edges (rounded in geometry). With todays computers (and my budget) that will not complete in the next years :wink:

This is no doubt the source of your splotches, and the photon map addition may well take care of the problem. You could try it with the "mirror" type first, just to see how long it takes and what it looks like.

Sounds like a field of a small application study of radiance and exotic model setup :wink: I have really little knowledge about the radiance pmap stuff, so it sound even more exciting. I was already thinking about putting an illum under the dome to achieve some smoothing, which would be more or less faked of course.

Thanks for the reply, I will keep you informed of any results...

CU Lars.