 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
No, this is not about the next version of POV-Ray..or maybe a little :)
I was looking at WinOSi homepage (http://www.winosi.onlinehome.de/) and
it seems to be a very interesting raytracing program. It is basically a
true forward raytracing engine: it shoots rays from light sources to the
scene - not the opposite as in many other renderes like POV-Ray, for
example. At least looking at the gallery it seems to be able to produce
some very realistic looking effects without any "faking" or tricks.
There is also no need to tweak tens of different settings to be able to
produce the desired output. You just set the scene and let the engine
render the image. The image quality gets better all the time as more
rays are calculated and it is up to the user to decide when the desired
quality has been reached.
The downside of all this is of course speed: it takes quite long for
WinOSi to produce good quality image even from a simple scene. But the
speed is not unbearable considering the quality and accuracy of the
images. Actually similar quality might be even slower to achieve using
backward raytracers.
I just began to wonder how other people thinks about the future of
POV-Ray. Has there been any thoughts of turning POV-Ray into a true
forward raytracer or are there issues (others than maybe rendering
speed) that makes it unfeasible? That might also make the code a lot
simpler as there would be no need for separate algorithms for photon
mapping, radiosity, anti-aliasing, specular highlights, color
dispersion, light sources, area/point lights etc. Not to mention the
simpler scene description files. WinOSi also seems to be very easily
parallelized. Or is the speed difference still so big that the extra
quality is not worth it?
BTW, are there forward renderes other than WinOSi available?
Severi S.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
sev### [at] NOT_THISsiba fi news:41954bb8$1@news.povray.org
> I just began to wonder how other people thinks about the future of
> POV-Ray. Has there been any thoughts of turning POV-Ray into a true
> forward raytracer or are there issues (others than maybe rendering
> speed) that makes it unfeasible? That might also make the code a lot
> simpler as there would be no need for separate algorithms for photon
> mapping, radiosity, anti-aliasing, specular highlights, color
> dispersion, light sources, area/point lights etc. Not to mention the
> simpler scene description files. WinOSi also seems to be very easily
> parallelized. Or is the speed difference still so big that the extra
> quality is not worth it?
IMHO it would a good idea to add this *possibility* and _not_ to limit
Pov-RAY to such approach - even now Pov-RAY is to slow for meany
profesional graphics because it doesnt offert line-scane mode.
1. line-scan mode (+ some tricks)
2. ray-tracing
3. light-tracing
If I'm correct (am I?) this are mose pupular approaches, currenlty PovRAY
implements 2.
Also, Pov-RAY even "traces" every shape, instead of working on triangles.
IMHO best way to go would be to increase aspect of Pov-RAY futures, to
allow both faster-but-faked renderings, and slow-but-accurate ones.
I myself had had some ideas for TurboPOV, for example - to pre-render
shadow (especialy hi-quality area_light) and use it later just as a
texture, same could be done for i.e. radiosity and photons.
Assume we hawe glass-ball +photons+dispersion over a plane. Then we pre-
trace the plane image, and use pre-traced data as image map. That way, if
other objects like reflective sphere wants to "read" color of plane, it
doesnt trigger again photon calculations, it only read's data from
image_map.
Other approach could be to pre-traingulate some compilated objects into
smooth-mashes (or at least use them as better bounding-objects).
I wonder if any of thoes ideas is worth implementing? It should give big
speed boost in some cases...
--
http://www.raf256.com/3d/
Rafal Maj 'Raf256', home page - http://www.raf256.com/me/
Computer Graphics
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Severi Salminen wrote:
>
> The downside of all this is of course speed: it takes quite long for
> WinOSi to produce good quality image even from a simple scene.
That's an understatement. Renderers taking this approach take extremely
long for any non trivial scene - a typical complex landscape would take
years (!!!) to render in a quality comparable to a POV-Ray render.
> BTW, are there forward renderes other than WinOSi available?
Such renderers are quite frequently used for generating reference images
in research - for example to demonstrate that a certain other technique
can generate accurate results much faster.
Christoph
--
POV-Ray tutorials, include files, Sim-POV,
HCR-Edit and more: http://www.tu-bs.de/~y0013390/
Last updated 23 Sep. 2004 _____./\/^>_*_<^\/\.______
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Christoph Hormann wrote:
>> The downside of all this is of course speed: it takes quite long for
>> WinOSi to produce good quality image even from a simple scene.
>
> That's an understatement. Renderers taking this approach take extremely
> long for any non trivial scene - a typical complex landscape would take
> years (!!!) to render in a quality comparable to a POV-Ray render.
What do you mean by "a typical complex landscape"? Textured isosurface +
media + global illumination? It would be nice to be able to make some
comparisons between different methods. I'm not sure if WinOSi supports
heightfields now (or media). It does support surfaces of the form
z=f(x,y). Of course, landscape images are only one small area of
raytracing. There are N other types of scenes where forward raytracing
is the only way to get correct results.
>> BTW, are there forward renderes other than WinOSi available?
>
> Such renderers are quite frequently used for generating reference images
> in research - for example to demonstrate that a certain other technique
> can generate accurate results much faster.
Do you happen to know any which have a webpage?
Severi
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Severi Salminen wrote:
>>
>> That's an understatement. Renderers taking this approach take
>> extremely long for any non trivial scene - a typical complex landscape
>> would take years (!!!) to render in a quality comparable to a POV-Ray
>> render.
>
>
> What do you mean by "a typical complex landscape"? Textured isosurface +
> media + global illumination?
No, i am talking about any scene that has the complexity of a typical
landscape scene. What kind of geometry representation you use does not
matter much here.
If you have a look at the WinOSi site - the only scene they give render
times for is about 20 minutes in POV-Ray and 40 hours with WinOSi.
That's about 120 times as long - a complex POV-Ray scene at print size
might render a week - that would be more than 2 years with the same
ratio but POV-Ray scales much better with higher complexity (due to
irradiance cache for example).
Christoph
--
POV-Ray tutorials, include files, Sim-POV,
HCR-Edit and more: http://www.tu-bs.de/~y0013390/
Last updated 23 Sep. 2004 _____./\/^>_*_<^\/\.______
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
In article <41954bb8$1@news.povray.org>,
Severi Salminen <sev### [at] NOT_THISsibafi> wrote:
> No, this is not about the next version of POV-Ray..or maybe a little :)
>
> I was looking at WinOSi homepage (http://www.winosi.onlinehome.de/) and
> it seems to be a very interesting raytracing program. It is basically a
> true forward raytracing engine: it shoots rays from light sources to the
> scene - not the opposite as in many other renderes like POV-Ray, for
> example. At least looking at the gallery it seems to be able to produce
> some very realistic looking effects without any "faking" or tricks.
> There is also no need to tweak tens of different settings to be able to
> produce the desired output. You just set the scene and let the engine
> render the image. The image quality gets better all the time as more
> rays are calculated and it is up to the user to decide when the desired
> quality has been reached.
It is not a "true forward raytracing engine". Such an engine would
simply be too slow, requiring huge amounts of time and processing power
to render simple images. One image was rendered recently with such an
algorithm. It combined the results of of many computers around the world
over several months to generate an image with 382 billion photons.
http://www.cpjava.net/soupImages/soup_one.png
http://www.cpjava.net/photonproj.html
WinOSi uses some kind of variant on photon mapping...that is, a forward
raytracing technique used to modify illumination calculations in a
traditional reverse raytracer. This appears to do the equivalent of many
passes of gathering a relatively small number of photons, then
ray-tracing an image to be combined into the final image.
In any case, this way is very slow, and it is not clear that it is any
faster than equally accurate rendering using photon mapping...I would
not expect that to be so, from the description of the algorithm. It
doesn't appear to do anything that can't be done with photon mapping. I
think this approach is probably a dead end. There are other techniques
that give extremely good results with far less rendering time.
--
Christopher James Huff <cja### [at] earthlinknet>
http://home.earthlink.net/~cjameshuff/
POV-Ray TAG: <chr### [at] tagpovrayorg>
http://tag.povray.org/
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
|
|
