Impressive! Even by combining what is essentially 3 days worth of rendering 
I can't get that degree of clarity. Will take a look at the source.

Mike


"fidos" <fid### [at] wanadoofr> wrote in message 
news:web.499dafe53e484721c114ae9e0@news.povray.org...
> Hello,
>
> Clearly a difficult scene.
> A 14h hours rendering on 4 cores (Q6600 2.4 GHz).
> I made some change to the settings to speed up the rendering.
> The rendering was done with a modified version of mcpov 0.0.5 (the choise 
> of the
> pixel to shoot is different in order to reduce the noise were it is more
> important).
>
> Regrads,
> Fidos.
>

Post a reply to this message

This is what I get if I combine your image with my previous post and two 
other renders.  Looking better but I think fidos' render has it beat.


"scott" <sco### [at] scottcom> wrote in message news:499d0f7e@news.povray.org...
>> Does everyone's separate images have random grain (truly random, and not
>> identical from machine to machine)?
>
> It should do (AIUI the RNG is seeded from the system time).  Often people
> render on multi-core machines by starting several instances of MCPOV and
> then averaging later, so I think it works.
>
> The problem I see is that posting JPEG versions here will lose some of the
> detail, you really need to post PNG, but my file is 1.4 MB... Anyway, here
> is a JPEG version with minimum compression after 14 hours, this scene 
> really
> is a tricky beast!
>
>

Post a reply to this message

Attachments:
Download 'radiosity_mcpov_combined.jpg' (258 KB)

Preview of image 'radiosity_mcpov_combined.jpg'

"fidos" <fid### [at] wanadoofr> wrote:
> Hello,
>
> Clearly a difficult scene.
> A 14h hours rendering on 4 cores (Q6600 2.4 GHz).

WOW! That's really impressive.

So as I understand it, the more random-grain individual images that are averaged
together (images from different users on different machines), the better and
more accurate the scene becomes? That's fascinating. Each of us could spend
maybe 2 hours of spare computer time, rendering our (grainy) versions, them
send them back to the original author for the final averaging step. That's
just--cool!

Ken W.

Post a reply to this message

Kenneth wrote:

> So as I understand it, the more random-grain individual images that are averaged
> together (images from different users on different machines), the better and
> more accurate the scene becomes? That's fascinating. Each of us could spend
> maybe 2 hours of spare computer time, rendering our (grainy) versions, them
> send them back to the original author for the final averaging step. That's
> just--cool!

But only if they all use a different random seed! Averaging the same
grainy image 10 times won't improve the results at all.

-- 
Severi Salminen
http://www.iki.fi/severi

Post a reply to this message

Among other things, Kenneth saw fit to write:

> So as I understand it, the more random-grain individual images that are
> averaged together (images from different users on different machines), the
> better and more accurate the scene becomes? That's fascinating.

That's how signal and noise in general work. The "real" image is the signal,
the grain is the noise. As you add samples (rendered images) the signal
intensity increases linearly with the number of samples N, while the noise
typically increases only with the square root of N (because its random
nature). So to reduce the relative intensity of the noise 10 times you'd
need 100 times more images.

-- 
light_source{9+9*x,1}camera{orthographic look_at(1-y)/4angle 30location
9/4-z*4}light_source{-9*z,1}union{box{.9-z.1+x clipped_by{plane{2+y-4*x
0}}}box{z-y-.1.1+z}box{-.1.1+x}box{.1z-.1}pigment{rgb<.8.2,1>}}//Jellby

Post a reply to this message

Severi Salminen <sev### [at] NOTTHISsaunalahtifiinvalid> wrote:
> Kenneth wrote:
>
> > So as I understand it, the more random-grain individual images that are averaged
> > together (images from different users on different machines), the better and
> > more accurate the scene becomes? That's fascinating. Each of us could spend
> > maybe 2 hours of spare computer time, rendering our (grainy) versions, them
> > send them back to the original author for the final averaging step. That's
> > just--cool!
>
> But only if they all use a different random seed! Averaging the same
> grainy image 10 times won't improve the results at all.
>

I was looking through MCPOV's online documentation, and it indicates that each
separate image render (as in, images to be averaged later) should use
"mc_rand_seed" to set a different rand sequence: "Initiates the random sequence
used by the montecarlo algorithm with the current date." By 'date,' I'm guessing
that it means not just the particular day, but down to the second or even finer
time resolution (based on the machine's system clock.) So, for example, if I
start a render at 2:30:37 PM on March 16, and someone else in my time zone
happens to start one at 2:30:38 PM on the same day, the two rendered images
will still be randomly different. Am I correct?

KW

Post a reply to this message

"Kenneth" <kdw### [at] earthlinknet> wrote:
> I was looking through MCPOV's online documentation, and it indicates that each
> separate image render (as in, images to be averaged later) should use
> "mc_rand_seed" to set a different rand sequence: "Initiates the random sequence
> used by the montecarlo algorithm with the current date." By 'date,' I'm guessing
> that it means not just the particular day, but down to the second or even finer
> time resolution (based on the machine's system clock.) So, for example, if I
> start a render at 2:30:37 PM on March 16, and someone else in my time zone
> happens to start one at 2:30:38 PM on the same day, the two rendered images
> will still be randomly different. Am I correct?
>
> KW

The seed is initialized with the C time(NULL) call that gives the current GMT
time in seconds from 1970. So I think you are correct.

Fidos

Post a reply to this message

Kenneth wrote:
> So as I understand it, the more random-grain individual images that are
> averaged together (images from different users on different machines), the
> better and more accurate the scene becomes? That's fascinating. Each of us
> could spend maybe 2 hours of spare computer time, rendering our (grainy)
> versions, them send them back to the original author for the final
> averaging step. That's just--cool!

If I had some time (hahahaha) I'd resurrect my renderfarm and put MCPov on
it. And you could join :)

But there's that time thing.

Post a reply to this message