rtcontrib -c

Dear all,

I am just looking the rtcontrib -c option in the hope that it might be useful for reducing the graininess in images or noise in readings.

Where I am stuck is this: How do I get reinhart.cal to produce
those multiple rays which rtcontrib -c would then sum up?

This is for sensor illuminance readings:

$ cat data/2photocells.pts |rtcontrib -h -I $opt -e MF:1 \
-f reinhart.cal -b rbin -bn Nrbins \
-o tmp/sensors/s%03d.dat -c 10 \
-m sky_glow -w testroom_whitesky1.oct
rtcontrib: warning - partial accumulation in final record

Is the idea to create those multiple sample rays in a pre-run with rtrace or rcalc, and feed them to rtcontrib instead of the .pts file?

Thanks

Axel

Hi Axel,

The only time i've used the -c option used is in genklemsamp | rtcontrib for sampling window input directions. The -c for genklemsamp option determines the number of rays produced per klems division. And the -c option in rtcontrib tells how many rays to accumulate and average per output record.

In your case, you want to accumulate more values per pixel to achieve a smoother result. There may be a way to create a tricked out vwrays/rtcontrib combo to do this. for example:

vwrays ... | awk '{for (i=0,i<100,i++) print $0}' | rtcontrib -c 100 ... (completely untested!)

Alternatively you could run your rtcontrib command several times producing several sets of bin renderings. And then use pcomb to average the results of corresponding bin renderings into a hopefully less grainy bin rendering.

However you could probably reduce the graininess of your images by spawning more rays AND decreasing the threshold at which rays are considered insignificant and aborted. The second part of that is often overlooked and I find vitally important for pleasing results with rtcontrib. Increasing -ad helps to a point, but if you don't decrease -lw then increasing -ad doesn't increase the number of rays actually traced.

I have a rules of thumb for setting -lw:
- to trace every ray spawned, -lw should be roughly 1/(ad^ab) - watch out, every ray may be too much!
- if you have settings that produce a nice result with rpict try dividing -lw by the number of sky divisions for rtcontrib.

I seen significant grainniness improvements by reducing -lw.

Andy

···

On Apr 9, 2010, at 3:58 AM, Axel Jacobs wrote:

Dear all,

I am just looking the rtcontrib -c option in the hope that it might be useful for reducing the graininess in images or noise in readings.

Where I am stuck is this: How do I get reinhart.cal to produce
those multiple rays which rtcontrib -c would then sum up?

This is for sensor illuminance readings:

$ cat data/2photocells.pts |rtcontrib -h -I $opt -e MF:1 \
-f reinhart.cal -b rbin -bn Nrbins \
-o tmp/sensors/s%03d.dat -c 10 \
-m sky_glow -w testroom_whitesky1.oct
rtcontrib: warning - partial accumulation in final record

Is the idea to create those multiple sample rays in a pre-run with rtrace or rcalc, and feed them to rtcontrib instead of the .pts file?

Thanks

Axel

_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general

Hi Andy,

I have a rules of thumb for setting -lw:
- to trace every ray spawned, -lw should be roughly 1/(ad^ab) - watch out, every ray may be too much!
- if you have settings that produce a nice result with rpict try dividing -lw by the number of sky divisions for rtcontrib.

I seen significant grainniness improvements by reducing -lw.

Thanks for that. It's interesting that you set -lw to 1/ad^ab, not just 1/ad. Will try this out. I also feel that the number of sky divisions should have an impact on -ad (especially for sunny skies), but I can't quite put my finger on it.

Cheers

Axel

Using -lw set to 1/(ad^ab) can be a bit extreme, but as I said you should set -lw no higher than 1/ad. The smaller the value, the less noisy your indirect result, but a bit of experimentation on time/quality trade-off is (as always) a good idea.

Cheers,
-Greg

···

From: Axel Jacobs <[email protected]>
Date: April 9, 2010 3:16:56 PM PDT

Hi Andy,

I have a rules of thumb for setting -lw:
- to trace every ray spawned, -lw should be roughly 1/(ad^ab) - watch out, every ray may be too much!
- if you have settings that produce a nice result with rpict try dividing -lw by the number of sky divisions for rtcontrib.

I seen significant grainniness improvements by reducing -lw.

Thanks for that. It's interesting that you set -lw to 1/ad^ab, not just 1/ad. Will try this out. I also feel that the number of sky divisions should have an impact on -ad (especially for sunny skies), but I can't quite put my finger on it.

Cheers

Axel