 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Hallo,
http://home.arcor.de/mad_onion/005%20rund%20-%20bett.jpg
As you can see, the room is artifact-free except the one marked
with the arrow. I often have mainly clean images but with a few
big artifacts. How do these come into existance? How can I
prevent such artifacts? Is there something wrong with my scene
or my radiosity parameters (see below)?
Thank you,
Manuel
current radiosity parameters:
#include "rad_def.inc"
global_settings {
radiosity {
Rad_Settings( Radiosity_Normal, off, off)
recursion_limit 5
brightness 2.0
}
}
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Manuel Kasten nous apporta ses lumieres ainsi en ce 2004-10-13 17:02... :
>Hallo,
>
>http://home.arcor.de/mad_onion/005%20rund%20-%20bett.jpg
>As you can see, the room is artifact-free except the one marked
>with the arrow. I often have mainly clean images but with a few
>big artifacts. How do these come into existance? How can I
>prevent such artifacts? Is there something wrong with my scene
>or my radiosity parameters (see below)?
>
>Thank you,
> Manuel
>
>
>
>current radiosity parameters:
>
>#include "rad_def.inc"
>global_settings {
> radiosity {
> Rad_Settings( Radiosity_Normal, off, off)
> recursion_limit 5
> brightness 2.0
> }
>}
>
>
>
>
Try lowering error_bound from it's default value of 1.8. Do it by 0.1 or
less decreases. In some cases, a small increase may give some
improvement. Also, increasing the count may help, default = 35.
Setting pretrace_end to 0.02 or 0.01 or less can aleviate some artifacts
in some cases (default = 0.04)
Alain
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
>>current radiosity parameters:
>>
>>#include "rad_def.inc"
>>global_settings {
>> radiosity {
>> Rad_Settings( Radiosity_Normal, off, off)
>> recursion_limit 5
>> brightness 2.0
>> }
>>}
>>
> Try lowering error_bound from it's default value of 1.8. Do it by 0.1 or less
decreases. In some cases, a small increase may give
> some improvement. Also, increasing the count may help, default = 35.
> Setting pretrace_end to 0.02 or 0.01 or less can aleviate some artifacts in some
cases (default = 0.04)
Thanks, but the definitions in rad_def.inc include all this
(and they are very good, as I really like the rest of the room...):
#case (Radiosity_Normal)
pretrace_start 0.08
pretrace_end 0.01
count 200
nearest_count 7
error_bound 0.3
recursion_limit 1
low_error_factor 0.75
gray_threshold 0
minimum_reuse 0.017
brightness 1.0
adc_bailout 0.01/2
#debug "\nRadiosity_Normal in use\n"
#break
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Manuel Kasten <kas### [at] gmx de> wrote:
> As you can see, the room is artifact-free except the one marked
> with the arrow.
The reason for that artifact is a rather simple one, but I can't think
of a good solution in the general case. In this specific case there's
probably a working solution.
The problem is that at that specific point (the center of the circle)
the radiosity samples are taken from a point which is almost exactly at
the corner of the room, ie. at the junction of the two walls.
That is, that point is practically in the same place as the wall on
the right.
Now, think how the radiosity sampling works: POV-Ray sends semi-random
rays from that point to all directions (well, in all directions of the
hemisphere pointing out of the wall surface).
Since the sampling point is located just at the junction of the two
walls, some of these sample rays miss the right wall and are actually
going outside the room. Since it's dark outside, it affects the lighting
of that point.
I really can't think how this problem could be fixed in the POV-Ray
source code.
However, you could try translating your camera by a minuscule amount
(like 0.001) to some direction (but not vertically, because the walls
are vertical at the problematic place), and it might correct the problem.
It's enough for the sampling point to be taken just slightly from a different
place.
If you are out of luck, you might get a similar artifact somewhere else.
If this is the case, try translating the camera by a different amount.
Another possible solution (which is not a generic one but which probably
works here) is to make the right wall shadowless and put a copy of it
behind it (very close to it, but not too close). This way the sample
rays going outside the room will hit this second wall (which will be
illuminated by the light source because the original wall is now
shadowless).
By the way, can I use this image (well, a section of it) to make a new
entry in the http://tag.povray.org/povQandT/ site explaining this problem?
--
#macro M(A,N,D,L)plane{-z,-9pigment{mandel L*9translate N color_map{[0rgb x]
[1rgb 9]}scale<D,D*3D>*1e3}rotate y*A*8}#end M(-3<1.206434.28623>70,7)M(
-1<.7438.1795>1,20)M(1<.77595.13699>30,20)M(3<.75923.07145>80,99)// - Warp -
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> The reason for that artifact is a rather simple one, but I can't think
> of a good solution in the general case. In this specific case there's
> probably a working solution.
>
> The problem is that at that specific point (the center of the circle)
> the radiosity samples are taken from a point which is almost exactly at
> the corner of the room, ie. at the junction of the two walls.
> That is, that point is practically in the same place as the wall on
> the right.
>
> Now, think how the radiosity sampling works: POV-Ray sends semi-random
> rays from that point to all directions (well, in all directions of the
> hemisphere pointing out of the wall surface).
> Since the sampling point is located just at the junction of the two
> walls, some of these sample rays miss the right wall and are actually
> going outside the room. Since it's dark outside, it affects the lighting
> of that point.
Thank you for this detailed description!
> I really can't think how this problem could be fixed in the POV-Ray
> source code.
> However, you could try translating your camera by a minuscule amount
> (like 0.001) to some direction (but not vertically, because the walls
> are vertical at the problematic place), and it might correct the problem.
> It's enough for the sampling point to be taken just slightly from a different
> place.
> If you are out of luck, you might get a similar artifact somewhere else.
> If this is the case, try translating the camera by a different amount.
Mhh. That's not what I hoped for. Rendering such an image takes more than an
hour, and (repeatedly) moving the camera until no artifact appears can be
frustrating.
> Another possible solution (which is not a generic one but which probably
> works here) is to make the right wall shadowless and put a copy of it
> behind it (very close to it, but not too close). This way the sample
> rays going outside the room will hit this second wall (which will be
> illuminated by the light source because the original wall is now
> shadowless).
Well, if the geometry of the "room" is simple, that could be a solution.
When I get home, I'll try if it has any sideeffects...
> By the way, can I use this image (well, a section of it) to make a new
> entry in the http://tag.povray.org/povQandT/ site explaining this problem?
Sure, no problem. If you need the original (as png without the arrow), mail
me.
Manuel
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Warp wrote:
> However, you could try translating your camera by a minuscule amount
> (like 0.001) to some direction (but not vertically, because the walls
> are vertical at the problematic place), and it might correct the problem.
> It's enough for the sampling point to be taken just slightly from a different
> place.
> If you are out of luck, you might get a similar artifact somewhere else.
> If this is the case, try translating the camera by a different amount.
>
> Another possible solution (which is not a generic one but which probably
> works here) is to make the right wall shadowless and put a copy of it
> behind it (very close to it, but not too close). This way the sample
> rays going outside the room will hit this second wall (which will be
> illuminated by the light source because the original wall is now
> shadowless).
>
If the problem is caused by the sharp angle, wouldn't it be possible to
smooth the edge, for example by fitting a quarter cylinder in the corner
? I've never tried that, but maybe it could solve the missing
intersection problem ?
Regards
--
Vincent
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Vincent LE CHEVALIER <lec### [at] ctiecpfr> wrote:
> If the problem is caused by the sharp angle, wouldn't it be possible to
> smooth the edge, for example by fitting a quarter cylinder in the corner
Yes, that probably solves the problem.
--
#macro M(A,N,D,L)plane{-z,-9pigment{mandel L*9translate N color_map{[0rgb x]
[1rgb 9]}scale<D,D*3D>*1e3}rotate y*A*8}#end M(-3<1.206434.28623>70,7)M(
-1<.7438.1795>1,20)M(1<.77595.13699>30,20)M(3<.75923.07145>80,99)// - Warp -
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Hallo again,
your description holds true for the green marked artifact in:
http://home.arcor.de/mad_onion/rad_art_001.jpg
(note: I lowered radiosity settings for faster testing)
But whats with the red marked? I don't see how this is possible...
Plus, re-rendering this picture, without changing anything, has
vanished both artifacts. Is this related to anti-alias-jitter?
Thanks for enlightening me,
Manuel
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
|
|
