 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Rafal 'Raf256' Maj wrote:
> I suggest 2 technical changes,
> 1. much stronger AA (and in addition - render image in double resolution
> and then resample down using some good[1] graphics program).
I think it is much more efficient to _just_ use stronger AA than to
render the image larger and resize (that is if you don't actually need
the bigger image). Adaptive AA is basically the same as doing a larger
render but is computationally cheaper.
Severi
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
sev### [at] NOT_THISsiba fi news:40921b0d$1@news.povray.org
> Adaptive AA is basically the same as doing a larger
> render but is computationally cheaper.
No, it doesnt.
Try rendering one of most aliasing patter - black&white checke (ambietn 1)
and try if You can rid of aliasing even with very strong aa
--
http://www.raf256.com/3d/
Rafal Maj 'Raf256', home page - http://www.raf256.com/me/
Computer Graphics
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Rafal 'Raf256' Maj wrote:
>>Adaptive AA is basically the same as doing a larger
>>render but is computationally cheaper.
>
>
> No, it doesnt.
>
> Try rendering one of most aliasing patter - black&white checke (ambietn 1)
> and try if You can rid of aliasing even with very strong aa
Umm, yes I can. Just try with "+a0.0 +am2 +r5" for starters - the
checkered plane looks quite good in the horizont at least with 640x480.
Then test how much bigger you have to render without AA and resize to
get same results - 6400x4800 was (obviously) not enough . And read the
description from manual. You won't see good results if you use too big
value with +an.n.
Severi
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
sev### [at] NOT_THISsibafi news:4093ca9a$1@news.povray.org
> get same results - 6400x4800 was (obviously) not enough . And read the
> description from manual. You won't see good results if you use too big
> value with +an.n.
I know what +an.n means, and I didnt told to render in higher resolution.
I suggested to render in double resolution *and* with strong AA.
--
http://www.raf256.com/3d/
Rafal Maj 'Raf256', home page - http://www.raf256.com/me/
Computer Graphics
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Rafal 'Raf256' Maj wrote:
>>get same results - 6400x4800 was (obviously) not enough . And read the
>>description from manual. You won't see good results if you use too big
>>value with +an.n.
>
>
> I know what +an.n means, and I didnt told to render in higher resolution.
> I suggested to render in double resolution *and* with strong AA.
Yes, and I said that you can achieve the same easier and even more
efficiently by just using stronger AA. There is no need for
enlarging+resizing to get rid of Aliasing.
Severi
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
de news: 4093cc9f@news.povray.org...
> Rafal 'Raf256' Maj wrote:
> Yes, and I said that you can achieve the same easier and even more
> efficiently by just using stronger AA. There is no need for
> enlarging+resizing to get rid of Aliasing.
>
Doesn't AA make the same thing?
computing more points just as if image size was bigger and interpolating
them just as an image editor when you reduce resolution?
Marc
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
jac### [at] wanadoofr news:4093e81c$1@news.povray.org
> computing more points just as if image size was bigger and interpolating
> them just as an image editor when you reduce resolution?
AA tries to do it in smart way (compute as mych extra rays - supersamples
per one final pixel as needed for this pixel), but sometimes this adaptive
method gives wrong results, therefore using it *and* double resoulution is
the fastest way to get rid of heavy aliasing.
--
http://www.raf256.com/3d/
Rafal Maj 'Raf256', home page - http://www.raf256.com/me/
Computer Graphics
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> AA tries to do it in smart way (compute as mych extra rays - supersamples
> per one final pixel as needed for this pixel), but sometimes this adaptive
> method gives wrong results, therefore using it *and* double resoulution is
> the fastest way to get rid of heavy aliasing.
What do you mean by "wrong results"? Could you please post a simple example
scene (and AA settings) where AA gives those wrong results? I'd be very
interest to see it.
Severi S.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
sev### [at] NOT_THISsibafi news:40940aaf$1@news.povray.org
> What do you mean by "wrong results"? Could you please post a simple
> example scene (and AA settings) where AA gives those wrong results?
> I'd be very interest to see it.
Hmm it seems that +am2 works exacly as resizing, *but* this behaviour is
not as it is described in docs. (+am2 results in non-adaptive antialiasing,
always maximum number of supersamples are used even if we are rendering an
empty black scene)
--
http://www.raf256.com/3d/
Rafal Maj 'Raf256', home page - http://www.raf256.com/me/
Computer Graphics
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
> Hmm it seems that +am2 works exacly as resizing, *but* this behaviour is
> not as it is described in docs. (+am2 results in non-adaptive
antialiasing,
> always maximum number of supersamples are used even if we are rendering an
> empty black scene)
I think it does follow the docs description - if not, you should maybe
report this at p.d.i. You can't see the adaptation with +a0.0 as all
subsamples are calculated - as written in docs. With non-zero threshold the
adaptation kicks in. I think that is correct functioning. If I misunderstood
your point, please corect me.
Severi S.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
