Apologies for the on-topic post...

Has anyone given any thought to altering the blur kernel in POVRay? 
Currently it's rectangular, but really should be a Gaussian, I believe, 
or at least allow for circular, with or without falloff.

What would be extra nice would be to add the ability to define number of 
  aperture blades, whether the blades are curved, and the "structure" of 
the bokeh (e.g. With some lenses, the circle of confusion is rather 
constant, others have a fall off toward the edges, while others get 
brighter at the edges (though, this is considered bad bokeh)) It would 
lend some nice realism to scenes that use focal blur, I think...

IIRC, someone had set up a scene to do just that a while back, actually 
doing it in SDL.

-- 
~Mike

Post a reply to this message

Mike Raiford <"m[raiford]!at"@gmail.com> wrote:
> Has anyone given any thought to altering the blur kernel in POVRay?

I'll be brief:


uses obstructed rays.

http://news.povray.org/povray.binaries.images/thread/%3Cweb.47bf5650425b71eaae42298f0



    while( ! in_circle )
         try_again



Shameless example of use:

http://news.povray.org/povray.binaries.images/thread/%3Cweb.47c26b5e8e18150dae42298f0




http://pov.monde.free.fr/mael/mlpov081eng.html


> IIRC, someone had set up a scene to do just that a while back, actually
> doing it in SDL.

Hey, I poke my head in for five minutes and my work comes up!  I therefore
conclude that I'm the topic of many lively discussions!  Please, I need this.
Really though, I apologize that the right convergence of desire and opportunity
never came up so that I could learn how to patch POV properly and make something
people can use.  Now isn't the time either--of course--but maybe it needs to be
done.

 - Ricky

Post a reply to this message

"triple_r" <nomail@nomail> wrote:

Sorry.

>
http://news.povray.org/povray.binaries.images/thread/%3Cweb.47bf5650425b71eaae42298f0

http://news.povray.org/povray.binaries.images/thread/%3Cweb.47bf5650425b71eaae42298f0%40news.povray.org%3E/?ttop=318663
&toff=750

>
http://news.povray.org/povray.binaries.images/thread/%3Cweb.47c26b5e8e18150dae42298f0

http://news.povray.org/povray.binaries.images/thread/%3Cweb.47c26b5e8e18150dae42298f0%40news.povray.org%3E/?ttop=318663
&toff=750

 - Ricky

Post a reply to this message

Mike Raiford schrieb:
> Apologies for the on-topic post...

You could have posted povray.programming (or, alternatively, 
povray.pov4.discussion.org), for instance...


> Has anyone given any thought to altering the blur kernel in POVRay? 
> Currently it's rectangular, but really should be a Gaussian, I believe, 
> or at least allow for circular, with or without falloff.

I presume you're talking about focal blur; well, it actually depends on 
the number of samples you have it take. Values of 7, 19, 37, or anything 
closely above these, will actually give you a hexagonal shape.

Yes, focal blur can definitely be improved. I don't think it ranks too 
high on the agenda though.

Post a reply to this message

clipka wrote:

> 
> I presume you're talking about focal blur; well, it actually depends on 
> the number of samples you have it take. Values of 7, 19, 37, or anything 
> closely above these, will actually give you a hexagonal shape.
> 

I just tried this, neat!

Curious, but why is this? Are there any higher sample numbers that give 
a hexagonal shape?

-- 
~Mike

Post a reply to this message

Mike Raiford schrieb:
> Curious, but why is this? Are there any higher sample numbers that give 
> a hexagonal shape?

No, these are the only ones (at least at present).

What POV-Ray actually does is use the largest of several hard-coded 
patterns (plus jitter) that fits the target count, then add some 
arbitrary points to get the target count.

The hard-coded patterns are a 4-sample pattern as well as a 7-sample, 
19-sample and 37-sample pattern. Except for the 4-sample pattern, all of 
these fit nicely on a triangular grid to form a hexagon, which is what 
gives you a hexagonal bokeh if you exactly match these numbers.

The additional random points are then taken from a square grid, which is 
what leads to the square bokeh typical for other sample counts.

Post a reply to this message