[Radiance-general] rendering meshes with interpolated vertex normals

[Greg Ward]

Mark is right.  To get the correct appearance out of a triangle mesh,  
you actually need to use a displacement map to give the underlying  
model the correct appearance w.r.t. grazing shadows such as this  
one.  Implementing displacement maps in a ray tracer is pretty  
difficult, and in the end it costs more to render them than simply  
increasing the resolution of your T-mesh as Mark suggests.

-Greg

> From: Mark Stock <mstock at umich.edu>
> Date: December 13, 2005 6:49:52 AM PST
>
> Bob,
>
> Your guess is right. Radiance's triangles, even with normal
> interpolation, are still flat. To use the normals to present a true  
> curved surface would (I think) require an implicit representation  
> of the surface.
>
> This particular example, with a low-res mesh and direct sunlight,  
> is the "poster-child" of this problem.
>
> I usually avoid this problem by a combination of higher resolution  
> (I use tri meshes up to 1M elements) and more ambient than direct  
> light.
>
> So, I see no way to completely avoid this for some problems.
>
> Mark
> mstock at umich.edu
>
> On Tue, 13 Dec 2005, bob coyne wrote:
>
>> Hi,
>>
>> I'm having a problem rendering polygon mesh objects with interpolated
>> vertex normals (converted via obj2mesh). I'm running Radiance 3.7.2b
>> patch release 22 Aug 2005.
>>
>> When I render a sphere mesh that's illuminated from above and to the
>> side, I get bad artifacts along the shadow line, with large dark
>> triangular wedges periodically protruding into the illuminated
>> area. (See link below for what this looks like.)  The dark triangles
>> seem to correspond to those quadrilaterals in the mesh where only a
>> single vertex is illuminated. My guess is that it's treating the  
>> whole
>> quad as pointing away from the light source, even though one vertex
>> and a significant area of it are in the illuminated region of the
>> object.
>>
>> Is there any way to avoid this?
>>
>> www.semanticlight.com/images/sphere-problem.jpg
>>
>> Thanks for any help.
>>
>> - Bob
>