|
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Hi,
I am trying to apply radiocity to my scene. Final image looks fine except
for some strange horizontal lines here and there.
I tried to adjust different parameters, almost without any result. It seems
that the only parameter which affects the appearence of these lines is
"error_bound". When it is below 2.0, the lines start to enumerate. Close to
0.01 they become so numerous that they form ugly patches. Above 2.0 there
is practically no difference.
What is wrong? Could someone please help?
Post a reply to this message
Attachments:
Download 'rad.jpg' (26 KB)
Preview of image 'rad.jpg'
|
|
| |
| |
|
|
|
|
| |
| |
|
|
avplat wrote:
> What is wrong? Could someone please help?
What does the rest of your radiosity block look like? Or could you post
the scene source?
I understand if you don't want to, but it might give some of us an
opportunity to play with the scene a bit to figure it out.
--
~Mike
Things! Billions of them!
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Mike Raiford wrote:
> What does the rest of your radiosity block look like? Or could you post
> the scene source?
>
> I understand if you don't want to, but it might give some of us an
> opportunity to play with the scene a bit to figure it out.
Radiosity block looks like this:
radiosity {
pretrace_start 0.08
pretrace_end 0.01
error_bound 3
gray_threshold 0.5
always_sample off
count 50
nearest_count 8
recursion_limit 3
}
Tim Nikias (I also posted this message in "newusers") suggested to play with
"nearset_count" and "count". My "nearest_count" is as big as he suggested.
Right now I am rendering with "count 300", but lines still remain, maybe a
little better. I will try 500 at night.
It is absolutely not a problem to post a source. But this is a kitchen with
all the furniture, etc. Original size is 27 MB... Archive size is around
1MB... It is not possible to post it here.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
"avplat" <nomail@nomail> wrote:
> Hi,
>
> I am trying to apply radiocity to my scene. Final image looks fine except
> for some strange horizontal lines here and there.
>
> I tried to adjust different parameters, almost without any result. It seems
> that the only parameter which affects the appearence of these lines is
> "error_bound". When it is below 2.0, the lines start to enumerate. Close to
> 0.01 they become so numerous that they form ugly patches. Above 2.0 there
> is practically no difference.
>
> What is wrong? Could someone please help?
It seems (to me) that the lines are being caused by your light sources.
try area lights to smooth the lines.
Wagner
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
avplat wrote:
> Mike Raiford wrote:
>
>
>>What does the rest of your radiosity block look like? Or could you post
>>the scene source?
>>
>>I understand if you don't want to, but it might give some of us an
>>opportunity to play with the scene a bit to figure it out.
>
>
> Radiosity block looks like this:
>
> radiosity {
> pretrace_start 0.08
> pretrace_end 0.01
> error_bound 3
> gray_threshold 0.5
> always_sample off
> count 50
> nearest_count 8
> recursion_limit 3
> }
>
> Tim Nikias (I also posted this message in "newusers") suggested to play with
> "nearset_count" and "count". My "nearest_count" is as big as he suggested.
> Right now I am rendering with "count 300", but lines still remain, maybe a
> little better. I will try 500 at night.
>
> It is absolutely not a problem to post a source. But this is a kitchen with
> all the furniture, etc. Original size is 27 MB... Archive size is around
> 1MB... It is not possible to post it here.
>
Bump up nearest_count and count. (Tim was certainly right), also turn
always_sample back on.
There *could* be a problem with coincident surfaces (I have seen very
strange lighting artifacts with them)
--
~Mike
Things! Billions of them!
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Mike Raiford wrote:
> Bump up nearest_count and count. (Tim was certainly right), also turn
> always_sample back on.
I tried with count 500. There are 3 lines remaining. I will try higher
parameter values, and will also try with different light types.
> There *could* be a problem with coincident surfaces (I have seen very
> strange lighting artifacts with them)
What do you mean?
What makes me wonder is that the lines are strictly horizontal related to
the *image*. They does not seem to relate to anything in the scene...
Btw, can it be antialiasing-related problem? I am using depth 3 with default
threshold (0.3).
Andrey
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
avplat wrote:
> Hi,
>
> I am trying to apply radiocity to my scene. Final image looks fine except
> for some strange horizontal lines here and there.
>
> I tried to adjust different parameters, almost without any result. It seems
> that the only parameter which affects the appearence of these lines is
> "error_bound". When it is below 2.0, the lines start to enumerate. Close to
> 0.01 they become so numerous that they form ugly patches. Above 2.0 there
> is practically no difference.
>
> What is wrong? Could someone please help?
>
>
>
>
> ------------------------------------------------------------------------
>
Hi there,
it may be strange but i had the same problem of brighter horizantal
lines and blocks also several times. But in my case it was not due to
radiosity but due to the choosen antialiasing parameters (as well on
win32 as on linux). Especialy when using method 1. Switching to method 2
normaly solved the problem in my cases.
(xcuse for my poor english)
dave
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
... and while we are talking about it,
could somebody explain me the fact that in some cases choosing
antialiasing method 1 in conjunction with low threshold limmits (<0.1)
casually produce banding (bright lines and blocks) in the rendered picture?
dave.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Hi,
After looking at numerous pre-rendered and rendered images again and again I
came to the conclusion, that these lines were not calculation errors of
some kind. Thus the problem could not be solved by adjusting radiosity
parameters.
This meant that Wagner perhaps was right. I started removing pieces of
furniture one by one - and inevitably found the one responsible for the
problem. It was a curved piece of glass - a part of the hood (not visible
at the image I presented). I am using Blender as a modelling tool, so I can
only use meshes, not the nice predifined POV-Ray objects. Chosen distance
between vertices was insufficient - which resulted in the surface being
slightly discontinuous. Rays of light went through this glass and formed
"bar code" pattern on the wall. Antialiasing further spoiled the picture by
removing most of these lines and leaving only some of them, making it
impossible to understand where they are coming from.
The problem was solved by increasing vertex count of the glass object and
its smoothing.
Thanks everybody for your advice.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
M-m-m, sorry, just noticed. I changed my pseudonym to real name, thus
"avplat" = "Andrey Platonov". :o)
--------------------------
Andrey
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |