fidos wrote:
> Severi Salminen <sev### [at] NOTTHISsaunalahtifiinvalid> wrote:
>> Severi Salminen wrote:
> It looks very well starting.

But I actually had a problem. I had a possibility to shoot more than 1
ray at each intersection. But I combined them incorrectly. I fixed that
and also noticed that it doesn't make sense to split the rays - one ray
at a time gave best results.

I also started using a faster way to create random rays and also
implemented importance sampling that made the image converge (smooth
out) faster. I attached a new example which _should_ be correct. Still
only balls but I have to figure the program structure first before
implementing more objects. I think that the basics should work now properly.

> I don't have source info. The idea is to help paths to find the light with fake
> objects. This objects are shooted in priority.

Hmm, I also read about a similar method where a ray is shot directly to
light sources at each bounce. Have to check it out.

Post a reply to this message

Attachments:
Download 'kuva4.jpg' (48 KB)

Preview of image 'kuva4.jpg'

Severi Salminen <sev### [at] NOTTHISsaunalahtifiinvalid> wrote:
> fidos wrote:
Your results are promising.

> I also started using a faster way to create random rays and also
> implemented importance sampling that made the image converge (smooth
> out) faster.
Seems interesting. What is importance sampling ? Do you have a reference paper ?

> Hmm, I also read about a similar method where a ray is shot directly to
> light sources at each bounce. Have to check it out.
The use of portals has the same effects when the are put around a light source.
But a portal can be put in the hall of a door, increasing the path finding
through the door. It can also be used on a refractive object to help to solve
caustic (but results are not always good in this case).

Regards,
Fidos.

Post a reply to this message

A new rendering. This scene is a demo of LightSys
(http://www.ignorancia.org/en/index.php?page=Lightsys).

I leaved it running 15 h on 4 cores of a Q6600.

Only brute force, no portal for this rendering (I will use it as reference image
for my next source modification of MegaPov).

Any comments.

Regards,
Fidos.

Post a reply to this message

Attachments:
Download 'demo_indoor1.04.jpg' (298 KB)

Preview of image 'demo_indoor1.04.jpg'

fidos napsal(a):
> A new rendering. This scene is a demo of LightSys
> (http://www.ignorancia.org/en/index.php?page=Lightsys).
> 
> I leaved it running 15 h on 4 cores of a Q6600.
> 
> Only brute force, no portal for this rendering (I will use it as reference image
> for my next source modification of MegaPov).
> 
> Any comments.
> 
> Regards,
> Fidos.
> 
> 
> 
> 
> ------------------------------------------------------------------------
> 
isn't the coach a bit too shiny?
-- 
the ultimate time-killer:
+a0.0 +am2 +r9

Johnny D

Post a reply to this message

A new rendering for only 30 minutes (on 4 cores) !
I don't anderstand why this scene rendered so quickly.
Good surprise.

The screne is from Gilles Tran
(http://www.oyonale.com/modeles.php?lang=en&page=37).

I used 1 portal around a spot light and a sky dome.

Regards,
Fidos.

Post a reply to this message

Attachments:
Download 'scifi_demo_.jpg' (395 KB)

Preview of image 'scifi_demo_.jpg'

> > I also started using a faster way to create random rays and also
> > implemented importance sampling that made the image converge (smooth
> > out) faster.
> Seems interesting. What is importance sampling ? Do you have a reference paper ?

It is really simple. Wikipedia also has an article about it. I also found some
ray tracing related articles but I don't now remember the source.

Normally you shoot rays in totally random directions inside a hemisphere and
weight the result according to cosine between normal and ray - in diffuse
reflections. (Mirrors are a different thing)

In importance sampling you don't shoot rays with uniform distribution. You shoot
them with cosine weighting (more rays to hemisphere pole, less to equator). And
because of this weighting you don't have to take the cosine into account
anymore. The real gain is that the image converges faster: you get more info
from interesting direction (normal). The result is still 100% unbiased.

See my thread in OT. Especially my last post where I realised that this works.
There is some wrong info also in the thread. My last post should be correct.

http://news.povray.org/povray.off-topic/thread/%3C47b81741%241%40news.povray.org%3E/

> The use of portals has the same effects when the are put around a light source.
> But a portal can be put in the hall of a door, increasing the path finding
> through the door. It can also be used on a refractive object to help to solve
> caustic (but results are not always good in this case).

Ok. I'll check that too. Bidirectional rendering looks interesting too.

Post a reply to this message

"fidos" <fid### [at] wanadoofr> wrote:
> A new rendering for only 30 minutes (on 4 cores) !
> I don't anderstand why this scene rendered so quickly.
> Good surprise.

No surprise at all. Path tracing works very well with large light sources. And
the largest one possible is usually sky. Most rays terminate very quickly and
the image converges very fast.

The slowest opposite is a unlit room with one single candle behind a corner - or
something similar. It is very random for a ray to actually reach and hit the
candle. Image will be noisy for a long time.

Post a reply to this message

amazing!  and almost no noise...

Post a reply to this message

Jan Dvorak nous apporta ses lumieres en ce 2008/02/20 15:23:

>> ------------------------------------------------------------------------
>>
> isn't the coach a bit too shiny?
In the original scene, it IS that shiny.

-- 
Alain
-------------------------------------------------
You know you've been raytracing too long when you've gained twenty pounds 
sitting at the computer, but can't tell because your beard covers your stomach.
Taps a.k.a. Tapio Vocadlo

Post a reply to this message

> No surprise at all. Path tracing works very well with large light sources. And
> the largest one possible is usually sky. Most rays terminate very quickly and
> the image converges very fast.
I agree with the theory. But the patio (a previuos screne that I posted in this
thread) was rendered with the same lighting (a portal around a point light and
a sky dome) and the rendering was far slower.

Regards,
Fidos

Post a reply to this message