This technique is known as 'monte-carlo ray-tracing' (used to be called
distributed ray-tracing) because of the randomness involved.

Somebody did implement a blury-reflections patch for POV.  You can find it
at the PatchStation... and I think it's part of the SuperPatch.
http://twysted.net/patchstation/

Blury refractions have not yet been implemented (afaik), but it would be
easy to extend the current patch to do this as you describe.

One thing:  most monte-carlo ray-tracers get information from an object's
BRDF to determine the amount of blur desired.  The BRDF (bi-directional
reflectance distribution function... or something like that) for POV is
made up of a combination of the diffuse, phong, specular, and reflect
components of the finish (along with the pigment).

For POV, though, I don't think it's necessary to use the BRDF to
determine the blur amount... a user-specifiable amount like you describe
would probably be adequate.

-Nathan

Steve wrote:
> 
> Fuzzy reflections can be acheived in the following way:
> 
> For a reflective object, define two new parameters called "blur" and another
> called "samples."  Blur's default is 0.0, with a maximum of 1.0.  Samples
> should be an integer number, with a default around 5 or so.
> When casting a reflected ray, instead of casting one ray cast "samples" number
> of rays and average their color values.  'Blur' should translate into an angle
> of deviation of the extra rays from the absolute reflected path..  Where 0.0
> means no deviation and 1.0 means the angles can deviate anywhere over the 180
> degrees of the hemisphere surrounding the surface normal at the point of
> intersection.  In order to avoid smearing and noise, a single ray should be
> cast in the absolute reflected direction, and this ray's value should way more
> heavily on the average than the other sample rays cast out.
> 
> Frosted glass is glass that refracts objects seen through them but in a
> fuzzier way.   Use the above method.  But use the refracted direction instead
> of the reflected direction and the anti-surface normal instead of the surface
> normal.  Also...changes in IOR between mediums are not necesarily needed for
> this to work. The results would look like transparent plastic or lampshades.
> 
> ----------
> Steve Horn

Post a reply to this message