"Jelle Duives" <jel### [at] tnonl> wrote:
> For a project we're using povray 3.7 rc6 to render a scene. For this scene (a
> scene with some houses, road & cars) we enabled radiosity in order to get more
> realistic images. However, when enabling radiosity, we observe a flickering
> effect of lights, a sort of 'disco' effect. Do you have any idea where this
> effect comes from and how to reduce this effect?
>
> Regards Jelle

clipka <ano### [at] anonymousorg> wrote:
> Am 05.09.2013 15:57, schrieb Jelle Duives:
> > For a project we're using povray 3.7 rc6 to render a scene. For this scene (a
> > scene with some houses, road & cars) we enabled radiosity in order to get more
> > realistic images. However, when enabling radiosity, we observe a flickering
> > effect of lights, a sort of 'disco' effect. Do you have any idea where this
> > effect comes from and how to reduce this effect?
>
> Radiosity isn't perfectly accurate, and may over- or underestimate the
> illumination in any given region of the image; if radiosity samples are
> not carried over from one frame to the next (which is the default), this
> may lead to visible random fluctuations in the brightness.
>
> If you don't have any moving elements in your scene (i.e. the animation
> is just a fly-through), this can be avoided by saving and restoring the
> radiosity samples from frame to frame.
>
> If you have stuff moving around, currently the only solution to the
> problem is to crank up the radiosity quality, to make the fluctuations
> as "shallow" as possible.

When rad is calculated each time and each scene is differently, even slightly,
it may come up with different results which causes the flickering effect.

The best way to handle this is to run a scene where you have a maximum view of
the entire scene (can even try to use some well placed mirrors to help) and run
with high radiosity settings and save the rad file.  You now should be able to
run the scene from your original view but load the rad file instead of
recalculating.

This should work better as the way rad is calculated and determined is dependent
upon the view when, but when it is save and reloaded, the view no longer matters
as it is mapped to the scene the same every time regardless of the new camera
view.  The only thing is areas that were hidden or occluded from the original
view may not get full rad effects which is why you want a setup that lets you
see as much of your scene as possible at one time.

I did a trick scene a long time ago (can't find it now) where I ran a rad scan
with a bunch of coloured balls, but then loaded the rad file with white balls
for an interesting colour cast effect.

-tgq

Post a reply to this message

"Trevor G Quayle" <Tin### [at] hotmailcom> wrote:
> "Jelle Duives" <jel### [at] tnonl> wrote:
> > For a project we're using povray 3.7 rc6 to render a scene. For this scene (a
> > scene with some houses, road & cars) we enabled radiosity in order to get more
> > realistic images. However, when enabling radiosity, we observe a flickering
> > effect of lights, a sort of 'disco' effect. Do you have any idea where this
> > effect comes from and how to reduce this effect?
> >
> > Regards Jelle
>
> clipka <ano### [at] anonymousorg> wrote:
> > Am 05.09.2013 15:57, schrieb Jelle Duives:
> > > For a project we're using povray 3.7 rc6 to render a scene. For this scene (a
> > > scene with some houses, road & cars) we enabled radiosity in order to get more
> > > realistic images. However, when enabling radiosity, we observe a flickering
> > > effect of lights, a sort of 'disco' effect. Do you have any idea where this
> > > effect comes from and how to reduce this effect?
> >
> > Radiosity isn't perfectly accurate, and may over- or underestimate the
> > illumination in any given region of the image; if radiosity samples are
> > not carried over from one frame to the next (which is the default), this
> > may lead to visible random fluctuations in the brightness.
> >
> > If you don't have any moving elements in your scene (i.e. the animation
> > is just a fly-through), this can be avoided by saving and restoring the
> > radiosity samples from frame to frame.
> >
> > If you have stuff moving around, currently the only solution to the
> > problem is to crank up the radiosity quality, to make the fluctuations
> > as "shallow" as possible.
>
> When rad is calculated each time and each scene is differently, even slightly,
> it may come up with different results which causes the flickering effect.
>
> The best way to handle this is to run a scene where you have a maximum view of
> the entire scene (can even try to use some well placed mirrors to help) and run
> with high radiosity settings and save the rad file.  You now should be able to
> run the scene from your original view but load the rad file instead of
> recalculating.
>
> This should work better as the way rad is calculated and determined is dependent
> upon the view when, but when it is save and reloaded, the view no longer matters
> as it is mapped to the scene the same every time regardless of the new camera
> view.  The only thing is areas that were hidden or occluded from the original
> view may not get full rad effects which is why you want a setup that lets you
> see as much of your scene as possible at one time.
>
> I did a trick scene a long time ago (can't find it now) where I ran a rad scan
> with a bunch of coloured balls, but then loaded the rad file with white balls
> for an interesting colour cast effect.
>
> -tgq

First of all thanks for your quick replies! I think that for our daylight
scenarios reusing the radiosity samples is a really good solution, we didn't
think of that yet. However, for our night scenes we have moving cars with lights
mounted 'on' them. In other words, we do not only have moving objects, but
moving lights as well.
Can we also reuse the radiosity samples for this type of scenarios? If not, is
there maybe another solution we can use?

Regards Jelle

Post a reply to this message

Jelle Duives wrote:

> If not, is there maybe another solution we can use?

The other solution is using higher quality radio settings,
and bringing a lot of patience (or computing power).

Post a reply to this message

>
> First of all thanks for your quick replies! I think that for our daylight
> scenarios reusing the radiosity samples is a really good solution, we didn't
> think of that yet. However, for our night scenes we have moving cars with lights
> mounted 'on' them. In other words, we do not only have moving objects, but
> moving lights as well.
> Can we also reuse the radiosity samples for this type of scenarios? If not, is
> there maybe another solution we can use?
>
> Regards Jelle
>
>
>
When you have moving parts, reusing radiosity samples will result in 
highly inacurate results. With moving lights, you'll be left with areas 
retaining the illumination from the first frame in the last frame.

Some things that you may try:

Use a smaller pretrace_end value. It default to 0.04. Try something like 
0.01 or 0.005.
Use a smaler value for low_error_factor. The default is 0.5. Try 
something around 0.2.
Use "+hr" on the command line to turn on the "high reproductibility" mode.
Increase the count value.
Take advantage of the importance feature: Set a relatively low default 
importance, but have critical objects with "radiosity{importance 1}" to 
use the maximum sampling for them. In your case, you should probably set 
the road surface and vehicles with a higher importance than the rest.
Sample of using importance:
#default{radiosity{ilportance 0.1}}
global_settings{radiosity{samples 1000}}
object{Some_Object{radiosity{importance 1}}

In this case, most of the scene will get about 100 radiosity samples, 
but "Some_object" will be sampled at 1000 samples. This can greatly 
reduce the trace time.



Alain

Post a reply to this message

Am 09.09.2013 07:15, schrieb Alain:

> Use "+hr" on the command line to turn on the "high reproductibility" mode.

Unfortunately that doesn't help with animations.

Post a reply to this message

> Am 09.09.2013 07:15, schrieb Alain:
>
>> Use "+hr" on the command line to turn on the "high reproductibility"
>> mode.
>
> Unfortunately that doesn't help with animations.
>
To bad. It would be nice if it did.

Post a reply to this message

Le 09/09/2013 07:15, Alain a écrit :
> When you have moving parts, reusing radiosity samples will result in
> highly inacurate results. With moving lights, you'll be left with areas
> retaining the illumination from the first frame in the last frame.
> 
> Some things that you may try:

Just a suggestion: flickering is high frequency, what about computing
(and saving) radiosity samples for "key" frames (such as every second or
something of that range for eye detection) and reusing the saved
radiosity on the next frames, until the next key frames.

On key frame, the radiosity is "exact" and it is stable for a few frames.
-- 
Just because nobody complains does not mean all parachutes are perfect.

Post a reply to this message

On Wed, 11 Sep 2013 08:20:36 +0200, Le_Forgeron <lef### [at] freefr> wro
te:


>> When you have moving parts, reusing radiosity samples will result in
>> highly inacurate results. With moving lights, you'll be left with are
as
>> retaining the illumination from the first frame in the last frame.
>>
>> Some things that you may try:
>
> Just a suggestion: flickering is high frequency, what about computing
> (and saving) radiosity samples for "key" frames (such as every second 
or
> something of that range for eye detection) and reusing the saved
> radiosity on the next frames, until the next key frames.
>
> On key frame, the radiosity is "exact" and it is stable for a few fram
es.

It would be nice if some kind of motion-blurr-like method could be used 
 

only on radiosity to smooth it out.

-- 

-Nekar Xenos-

Post a reply to this message

Nekar Xenos wrote:

> It would be nice if some kind of motion-blurr-like method could be used 
> only on radiosity to smooth it out.

You could render each frame twice (with radiosity data from previous
and next key frame and perform a weighted average later).

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> If you don't have any moving elements in your scene (i.e. the animation
> is just a fly-through), this can be avoided by saving and restoring the
> radiosity samples from frame to frame.

"Trevor G Quayle" <Tin### [at] hotmailcom> wrote:
> The best way to handle this is to run a scene where you have a maximum view of
> the entire scene (can even try to use some well placed mirrors to help) and run
> with high radiosity settings and save the rad file.  You now should be able to
> run the scene from your original view but load the rad file instead of
> recalculating.
>
> This should work better as the way rad is calculated and determined is dependent
> upon the view when, but when it is save and reloaded, the view no longer matters
> as it is mapped to the scene the same every time regardless of the new camera
> view.  The only thing is areas that were hidden or occluded from the original
> view may not get full rad effects which is why you want a setup that lets you
> see as much of your scene as possible at one time.
>
> I did a trick scene a long time ago (can't find it now) where I ran a rad scan
> with a bunch of coloured balls, but then loaded the rad file with white balls
> for an interesting colour cast effect.
>
> -tgq

This thread was helpful to me. I ended up using clipka's answer, but Trevor's
suggestion sounded very difficult - especially given the geometry of my scene.

So I enabled all three options documented here <
http://www.povray.org/documentation/3.7.0/r3_2.html#r3_2_8_8_2 > (i.e.
radiosity_file_name="my_file.rad_cache" radiosity_from_file=on
radiosity_to_file=on) and began a full animation render without trying to obtain
a nice radiosity cache first.

Interestingly, the first few frames showed turbulence, but they stabilized
immediately. This is probably fine for my purposes, because it's for a homework
art project and I'm restraining my inner quality stickler - and also because I'm
probably going to do a fade-in when I edit the video, so the first few frames
won't be very visible.

Now that I have the radiosity cache, I could surely rerender the first few
frames. But like I said, I'm trying not to be a stickler on this particular
project.

Post a reply to this message