Hi folks,

Anyone interested in test-driving a few unofficial additions to POV-Ray 
3.7.0, please have a look at https://github.com/UberPOV/UberPOV.

The most notable change is the addition of BLURRED REFLECTIONS, as well 
as a stochastic anti-aliasing (and generic oversampling) mode to go 
along with it.


Two other unrelated changes are also included:

One is the ability to read arbitrary plaintext lines from ASCII files.

The other applies to overexposed colors in non-HDRI output images; a new 
option has been added to boost any non-overexposed color channel(s) 
(effectively white-shifting such colors) to compensate for the clipping 
of the overexposed channel(s), in an attempt to preserve brightness.


The features are documented in changes.txt.

A Windows binary is not available yet.


More is yet to come. An obvious next step will be to implement a Monte 
Carlo-based replacement for radiosity (I guess you know where this is 
heading), but a few odds & ends will also be thrown in, such as a 
feature to determine the modification time of a file (so that you can 
e.g. decide whether some auto-generated files might be outdated).


Please use povray.unofficial.patches for any feedback.

Post a reply to this message

> Hi folks,
>
> Anyone interested in test-driving a few unofficial additions to
> POV-Ray 3.7.0, please have a look at
> https://github.com/UberPOV/UberPOV.
>
> The most notable change is the addition of BLURRED REFLECTIONS, as
> well as a stochastic anti-aliasing (and generic oversampling) mode to
> go along with it.

   How convenient that I had no time this weekend to continue with the
baubles scene... :)


P.S.: I was really missing the smell of an unofficial patch in the
morning...

--
jaime

Post a reply to this message

In povray.general clipka <ano### [at] anonymousorg> wrote:
> The most notable change is the addition of BLURRED REFLECTIONS

What are the advantages compared to the current trick (which can be
used to achieve images like http://iki.fi/warp/pics/Rubiks_Revenge2.jpg )?

-- 
                                                          - Warp

Post a reply to this message

Am 18.11.2013 15:36, schrieb Warp:
> In povray.general clipka <ano### [at] anonymousorg> wrote:
>> The most notable change is the addition of BLURRED REFLECTIONS
>
> What are the advantages compared to the current trick (which can be
> used to achieve images like http://iki.fi/warp/pics/Rubiks_Revenge2.jpg )?

There are three drawbacks of that trick:

(1) It is well known to scale poorly in terms of render time when any 
other oversampling features are involved, such as focal blur, area 
lights, and most notably other objects in the scene that also have 
blurred reflections.

(2) It is a mess to parameterize for physical consistency with the 
highlights.

(3) It becomes an ugly hassle when working with multi-layered (or 
otherwise complex) textures; this is not only an issue of ease of use, 
but also of render time. (Averaging multi-layered textures that only 
differ in the normals of the topmost layer is anything but effective, as 
not only the reflection but each other layer is computed N-fold as well.)

I know that you love that trick, and have in some tutorial made claims 
that it is a sufficient replacement for inbuilt blurred reflections, but 
the above facts make it no more than a nasty kludge. One that can be 
used to good effect if you invest sufficient time and effort, but a 
kludge nonetheless.


While I can't claim the current UberPOV implementation of blurred 
reflections to be particularly speedy, there are mechanisms in place to 
limit the number of tertiary rays when combining it with focal blur, 
area lights, subsurface light transport, or just a vast number of 
objects with blurred reflections, so that it doesn't get too bad with 
more complex scenes. There is also room to add some tweakables that may 
help speed things up.

The parameterization is designed such that using the same parameters as 
for the specular highlights will give physically accurate results. 
(Admittedly this only holds strictly true if you don't use variable 
reflection or fresnel, but that's not a problem with the reflection but 
with the current limitations of highlights; some future UberPOV version 
will address this as well.)

As for ease of use, it's as simple as adding a roughness parameter to 
the reflection block, and activating some feature that provides for 
sufficient oversampling (either focal blur, the new anti-aliasing mode, 
or both).

Post a reply to this message

Using VS2010 I get 3 compile time errors

1>..\..\source\backend\lighting\photons.cpp(417): error C2661: 
'pov::Trace::WNRX::WNRX' : no overloaded function takes 4 arguments

3>..\..\vfe\vfedisplay.cpp(61): error C2661: 
'pov_frontend::Display::Display' : no overloaded function takes 4 arguments

3>C:\Work\POV\UberPov\libraries\boost\boost/bind/bind.hpp(515): error 
C2593: 'operator [' is ambiguous

and as a result a link error as povbackend64.lib is not build.

Official POV 3.7 works.


As a side note: I'm curious to see how adaptive multi-level oversampling 
works out. To be honest, I'm considering to switch to a commercial 
render engine (I did already use the trial versions of the Maxwell and 
VRay renderer), not because of better quality but because both are so 
much faster when working with more complex scenes where e.g. things 
happen like blurred inter-reflections.
IMO the way POV-Ray implements various independent multi-sampling 
methods for various features (media, area_lights, focal blur and so 
on..) is the culprit here.

-Ive

Post a reply to this message

Am 18.11.2013 17:31, schrieb Ive:
>
> Using VS2010 I get 3 compile time errors
>
> 1>..\..\source\backend\lighting\photons.cpp(417): error C2661:
> 'pov::Trace::WNRX::WNRX' : no overloaded function takes 4 arguments
>
> 3>..\..\vfe\vfedisplay.cpp(61): error C2661:
> 'pov_frontend::Display::Display' : no overloaded function takes 4 arguments
>
> 3>C:\Work\POV\UberPov\libraries\boost\boost/bind/bind.hpp(515): error
> C2593: 'operator [' is ambiguous
>
> and as a result a link error as povbackend64.lib is not build.

The first two errors should be fixed now. Need to figure out what's 
wrong with the 3rd though.


> As a side note: I'm curious to see how adaptive multi-level oversampling
> works out. To be honest, I'm considering to switch to a commercial
> render engine (I did already use the trial versions of the Maxwell and
> VRay renderer), not because of better quality but because both are so
> much faster when working with more complex scenes where e.g. things
> happen like blurred inter-reflections.
> IMO the way POV-Ray implements various independent multi-sampling
> methods for various features (media, area_lights, focal blur and so
> on..) is the culprit here.

That's what I think, too.

I don't yet see the performance I was expecting, but I guess there's 
still quite some headroom for tweaking and optimizing.

Post a reply to this message

Am 18.11.2013 17:54, schrieb clipka:
> Am 18.11.2013 17:31, schrieb Ive:
>>
>> Using VS2010 I get 3 compile time errors
>>
>> 1>..\..\source\backend\lighting\photons.cpp(417): error C2661:
>> 'pov::Trace::WNRX::WNRX' : no overloaded function takes 4 arguments
>>
>> 3>..\..\vfe\vfedisplay.cpp(61): error C2661:
>> 'pov_frontend::Display::Display' : no overloaded function takes 4
>> arguments
>>
>> 3>C:\Work\POV\UberPov\libraries\boost\boost/bind/bind.hpp(515): error
>> C2593: 'operator [' is ambiguous
>>
>> and as a result a link error as povbackend64.lib is not build.
>
> The first two errors should be fixed now. Need to figure out what's
> wrong with the 3rd though.

Looks like some files didn't make it into the Git repo. Shoot me if it 
doesn't work NOW.

Post a reply to this message

Am 18.11.2013 17:54, schrieb clipka:
> Am 18.11.2013 17:31, schrieb Ive:
>>
>> Using VS2010 I get 3 compile time errors
>>
>> 1>..\..\source\backend\lighting\photons.cpp(417): error C2661:
>> 'pov::Trace::WNRX::WNRX' : no overloaded function takes 4 arguments
>>
>> 3>..\..\vfe\vfedisplay.cpp(61): error C2661:
>> 'pov_frontend::Display::Display' : no overloaded function takes 4
>> arguments
>>
>> 3>C:\Work\POV\UberPov\libraries\boost\boost/bind/bind.hpp(515): error
>> C2593: 'operator [' is ambiguous
>>
>> and as a result a link error as povbackend64.lib is not build.
>
> The first two errors should be fixed now. Need to figure out what's
> wrong with the 3rd though.

see povray.unofficial.patches for the follow-up.

Post a reply to this message

Am 18.11.2013 19:09, schrieb clipka:
>
> Looks like some files didn't make it into the Git repo. Shoot me if it
> doesn't work NOW.
>
No need for violence - successfully build.
Will play with it tomorrow and... thanks!

-Ive

Post a reply to this message

Ok, something goes wrong.
While the example scenes do work well a very simple scene of mine (where 
I just did try to add a bit of blurred reflection to a table) did not!
So I used one of your own scene file (stochastic_reflection_blur.pov) 
and simplified it until it shows the same problem.

command line:
+w800 +h600 +am2 +a0.1 +r3 +FE

None of the fancy features are even used anymore and render time is 
about 10 seconds with 3.7 but UberPov did after 10 minutes not even 
finish the first block.


//=================================================================================

//#version 3.7;

#version unofficial patch 3.7;
#patch "upov-reflection-roughness" 0.9;


global_settings {
   max_trace_level 25
   assumed_gamma 1.0
}

camera {
   perspective angle 50
   location  <0.0, 3,-5.0>
   right     x*image_width/image_height
   look_at   <0.0, 1.0, 0.0>
}

light_source {
   <3000,3000,0> color rgb 1
   area_light x*500,y*500, 9,9 adaptive 1 circular orient
}

sky_sphere {
   pigment {
     gradient <0,1,0>
     color_map {
       [0.00 srgb <0.6,0.7,1.0>]
       [0.35 srgb <0.1,0.0,0.8>]
       [0.65 srgb <0.1,0.0,0.8>]
       [1.00 srgb <0.6,0.7,1.0>]
     }
     scale 2
   }
}

plane{ <0,1,0>, 0
   texture{
     pigment{ checker color rgb 1 color rgb 0.1 }
     finish {
       ambient 0
       diffuse 0.7
       reflection { 1.0 fresnel} // roughness 0.005 }
   //    specular albedo 0.1
   //    roughness 0.005
     }
   }
   interior { ior 1.5 }
}


//=================================================================================

Post a reply to this message