Hello everybody.  :o)

I see you're currently discussing radiosity, so please let me add my
thoughts. I'm aware that the attached pictures aren't beautiful, but
they are meant as a suitable test. Hope you can accept them.

I experimented with some ideas, cause I look forward to the day where
raytracing produce realistic looking pictures, and I'm trying to make
POV do it, do it first, and do it BEST! Of course.

Please understand the concept: I have skipped all "light_source"
commands because they are poor fakes. Instead I try to use a realistic
enviroment. This means, I set up any object and let it's ambient value
determine how much light it emits with radiosity. Then I (as with the
sphere in the attached pictures) use blurred reflection to get specular
highlights on light-absorbing objects.. This is the realistic way of
making highlights, instead of specular / blinn / phong keywords.

I managed to balance these 3 things well:

(1) The "brightness" value in radiosity global settings.
(2) The strength of ambient values (going much over 1) to emit light.
(3) The diffuse & reflection keywords on light-receiving objects.

But the problem: I experimented with the best radiosity-settings I could
think out, and find.. With recursion_limit set to 1, results look pretty
good; actually very similar to ordinary light_sources but with some
plusses: a WHOLE object emit light, specular reflections are realistic,
ans render times acceptable.

BUT the light doesn't bounce off, more than once. It goes from the light
sources (the ambient objects) to the first receiving object, and no
further. So a shadow will still be completely dark, if there is no
direct light source on the other side.

The solution is obvious: Raise recursion_limit to 2, but I got stuck
here, as things immediately look awful, no matter what I do.. Does
anyone have an idea WHY, and is there a solution??  I would be
interested.. Unfortunately, render times on the awful picture was very
high.. I tried to further raise radiosity quality to get rid of the
artifacts, and had to break a rendering after 12 hours, not finished,
and the result didn't look any better.

So help me! Help all of us, to get realistic lighting! And please don't
say radiosity is great, until these big problems are solved.


Hugo

Post a reply to this message

Attachments:
Download 'realistic_light4.jpg' (58 KB) Download 'realistic_light7.jpg' (20 KB)

Preview of image 'realistic_light4.jpg'

Preview of image 'realistic_light7.jpg'

Hugo wrote:
> 
> [...]
> 
> Please understand the concept: I have skipped all "light_source"
> commands because they are poor fakes. Instead I try to use a realistic
> enviroment. This means, I set up any object and let it's ambient value
> determine how much light it emits with radiosity. Then I (as with the
> sphere in the attached pictures) use blurred reflection to get specular
> highlights on light-absorbing objects.. This is the realistic way of
> making highlights, instead of specular / blinn / phong keywords.
> 

I doubt there is much use in this, since a lot of other things are still
fake.  At least you would have to use photons too then (of course they
only work with light sources).

radiosity only effects the diffuse lighting of the scene, using extremely
high ambient objects is not necessarily more realistic than light
sources.  

> I managed to balance these 3 things well:
> 
> (1) The "brightness" value in radiosity global settings.
> (2) The strength of ambient values (going much over 1) to emit light.
> (3) The diffuse & reflection keywords on light-receiving objects.
> 
> But the problem: I experimented with the best radiosity-settings I could
> think out, and find.. With recursion_limit set to 1, results look pretty
> good; actually very similar to ordinary light_sources but with some
> plusses: a WHOLE object emit light, specular reflections are realistic,
> ans render times acceptable.

If they are while you use blurred reflection, your radiosity settings are
probably not very high quality.  The problems in the second pict look very
much like ordinary artefacts you could eliminate by increasing count and
tweaking some other values.  

> 
> BUT the light doesn't bounce off, more than once. It goes from the light
> sources (the ambient objects) to the first receiving object, and no
> further. So a shadow will still be completely dark, if there is no
> direct light source on the other side.
> 
> The solution is obvious: Raise recursion_limit to 2, but I got stuck
> here, as things immediately look awful, no matter what I do.. Does
> anyone have an idea WHY, and is there a solution??  I would be
> interested.. Unfortunately, render times on the awful picture was very
> high.. I tried to further raise radiosity quality to get rid of the
> artifacts, and had to break a rendering after 12 hours, not finished,
> and the result didn't look any better.
> 

That's what you have to expect.  Try using good old faked highlights and
things will be much faster.  Using light sources instead of pure rad
lighting usually also produces less artefacts.  

> So help me! Help all of us, to get realistic lighting! And please don't
> say radiosity is great, until these big problems are solved.
> 

You always have to remember that radiosity is not a all purpose fool proof
solution for all lighting problems.  I think it can help to achieve really
good results, but of course there are problems and disadvantages like slow
rendering, artefacts and like all other approximations (in fact everything
in Povray is faked), it has it's serious limits.

Christoph

-- 
Christoph Hormann <chr### [at] gmxde>
IsoWood include, radiosity tutorial, TransSkin and other 
things on: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message

> (in fact everything in Povray is
> faked), it has it's serious limits.

Shhh! <whisper>POV doesn't know that.</whisper> ;)

Post a reply to this message

Could you post your radiosity settings?

  More precisely: What is your error_bound?
  It may come as a bit of surprise, but try _increasing_ it (eg. to something
like 1 if you had it much smaller than that). You'll probably get a much
smoother illumination.

-- 
char*i="b[7FK@`3NB6>B:b3O6>:B:b3O6><`3:;8:6f733:>::b?7B>:>^B>C73;S1";
main(_,c,m){for(m=32;c=*i++-49;c&m?puts(""):m)for(_=(
c/4)&7;putchar(m),_--?m:(_=(1<<(c&3))-1,(m^=3)&3););}    /*- Warp -*/

Post a reply to this message

Warp wrote:
>   Could you post your radiosity settings?
>   More precisely: What is your error_bound?

Thanks for your interest to help!  :o)  Your suggestion is a surprise.
The settings I've used the most, are:

radiosity { pretrace_start .16 pretrace_end .02
            count 700 nearest_count 5 error_bound .2
            recursion_limit 2 low_error_factor .5 gray_threshold 0
            minimum_reuse 0.015 brightness 1 normal on }

This is with recursion_limit 2, and gives the yellow picture you saw. I
also tried to set the count to 1400, error_bound to 0.1, and
pretrace_end to 0.01 but it just takes very long to render, and looks
just as awful. On the other hand, the settings work nicely with
recursion set to 1. I wonder why.

Hugo

Post a reply to this message

Christoph Hormann wrote:
> radiosity only effects the diffuse lighting of the scene, using
> extremely high ambient objects is not necessarily more realistic
> than light sources.

The enviroment is closer to real life, this way. With these experiments
I tried to make POV step up on a higher level.. The attached pictures
were simple, and I compared them to a similar picture, lit by an
ordinary light_source - that actually looked more fake, IMO.

> Try using good old faked highlights and things will be much faster.

Faster, true. And I wanted to make a macro/include file that makes it
possible to setup the whole scene with normal light_soucres and just
shift to radlighting in the final rendering (automatically changing).

Someone have to break the barriers, and get raytracing to look better.
And I know it will happen soon. But maybe it will happen to big
commercial programs and not POVray. Who's gonna do it, if not ourselves?
We are all sort of responsible for POVs further development.

Thanks for your input.  :o)


Hugo

Post a reply to this message

Hugo wrote:
> 
> [...]
> 
> Someone have to break the barriers, and get raytracing to look better.
> And I know it will happen soon. But maybe it will happen to big
> commercial programs and not POVray. Who's gonna do it, if not ourselves?
> We are all sort of responsible for POVs further development.
> 

But don't forget one thing: Povray is a raytracer and therefore light
sources are an essential functionality.  IMO the things you consider as
more realistic are a quite subjective selection.  In some aspects blurred
reflection is much more fake than blinn highlights, furthermore as
mentioned before radiosity can only simulate the diffuse lighting of the
scene.  

Also the povray radiosity model was developed mainly for combination with
light sources.  In some way raytracing and radiosity are contradicting
models, a real (non monte carlo) radiosity program also needs to tesselate
all objects.

Referring to the things quoted above, it would be interesting to hear what
concrete things you have in mind that should be improved.

BTW, concerning your radiosity settings: 'adc_bailout' should be modified
when using high ambient objects.  The megapov docu suggests 
'0.01 / brightest_ambient_object'

Christoph

-- 
Christoph Hormann <chr### [at] gmxde>
IsoWood include, radiosity tutorial, TransSkin and other 
things on: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message

i don't think tweaking parameters in any program is a good way to go
about getting more realistic images.  in my opinion, the search for
realistic image making lies in better algorithms 'inside' the given
application.  beside, once you get settings you like with the current
pov, the next pov comes out and may change the underlying code you
relied on.  i really don't mean to discourage you.  your overall goal is
a good one; photorealism.

Hugo wrote:
> 
> Hello everybody.  :o)
> 
> I see you're currently discussing radiosity, so please let me add my
> thoughts. I'm aware that the attached pictures aren't beautiful, but
> they are meant as a suitable test. Hope you can accept them.
> 
> I experimented with some ideas, cause I look forward to the day where
> raytracing produce realistic looking pictures, and I'm trying to make
> POV do it, do it first, and do it BEST! Of course.
> 
> Please understand the concept: I have skipped all "light_source"
> commands because they are poor fakes. Instead I try to use a realistic
> enviroment. This means, I set up any object and let it's ambient value
> determine how much light it emits with radiosity. Then I (as with the
> sphere in the attached pictures) use blurred reflection to get specular
> highlights on light-absorbing objects.. This is the realistic way of
> making highlights, instead of specular / blinn / phong keywords.
> 
> I managed to balance these 3 things well:
> 
> (1) The "brightness" value in radiosity global settings.
> (2) The strength of ambient values (going much over 1) to emit light.
> (3) The diffuse & reflection keywords on light-receiving objects.
> 
> But the problem: I experimented with the best radiosity-settings I could
> think out, and find.. With recursion_limit set to 1, results look pretty
> good; actually very similar to ordinary light_sources but with some
> plusses: a WHOLE object emit light, specular reflections are realistic,
> ans render times acceptable.
> 
> BUT the light doesn't bounce off, more than once. It goes from the light
> sources (the ambient objects) to the first receiving object, and no
> further. So a shadow will still be completely dark, if there is no
> direct light source on the other side.
> 
> The solution is obvious: Raise recursion_limit to 2, but I got stuck
> here, as things immediately look awful, no matter what I do.. Does
> anyone have an idea WHY, and is there a solution??  I would be
> interested.. Unfortunately, render times on the awful picture was very
> high.. I tried to further raise radiosity quality to get rid of the
> artifacts, and had to break a rendering after 12 hours, not finished,
> and the result didn't look any better.
> 
> So help me! Help all of us, to get realistic lighting! And please don't
> say radiosity is great, until these big problems are solved.
> 
> Hugo
> 
>   ------------------------------------------------------------------------
>  [Image]  [Image]

Post a reply to this message

Hugo <hua### [at] post3teledk> wrote:
: radiosity { pretrace_start .16 pretrace_end .02
:             count 700 nearest_count 5 error_bound .2
:             recursion_limit 2 low_error_factor .5 gray_threshold 0
:             minimum_reuse 0.015 brightness 1 normal on }

  As I said, try a higher error_bound. I would try something like:

radiosity
{ pretrace_start .08
  pretrace_end .02
  count 200
  error_bound 1
  recursion_limit 2
  normal on
}

  Just try that and post the result.

-- 
char*i="b[7FK@`3NB6>B:b3O6>:B:b3O6><`3:;8:6f733:>::b?7B>:>^B>C73;S1";
main(_,c,m){for(m=32;c=*i++-49;c&m?puts(""):m)for(_=(
c/4)&7;putchar(m),_--?m:(_=(1<<(c&3))-1,(m^=3)&3););}    /*- Warp -*/

Post a reply to this message