sky mapping

Dear All,

I am trying to remap accurately an angular fisheye view of the sky using
the proj_pict_fisheye.cal developed by Peter Apian-Bennewitz, ISE, on
July 96 (as it is reported on the copyright header in the .cal file) to
subsequently use it as light source for a diffuse sky simulation.

It seems to work fairly well (the error at the centre is about 1% of the
original image) but

1) a bright circle forms along the horizon

2) the remapped image is constantly less bright than the original one

If anybody has any idea about the reason behind this behaviour it would
be very helpful

Thanks in advance,

Best Regards,

Giovanni

Hi!

It seems to work fairly well (the error at the centre is about 1% of
the original image) but

1) a bright circle forms along the horizon

2) the remapped image is constantly less bright than the original one

It would be nice if you could share the scene as well as the resulting
image by e.g. uploading it to a web oder ftp server. Please not by
posting any attachments.. If you have a bright ring in the result - do
you have a bright ring in your source image, e.g. at the edge of the
image circle? How did you prepare the image, do you crop it? Does the
cal-file expect a 180 fov. and does your image provide that?

Under 2) you write that the image is less bright then the original, but
at the centre you have only an error of 1% - do you have a constant
error of 1%?

Are you sure that the fisheye projection assumed by the script assumes
your camera/lens projection?

Does the source image include direct sun (visible)?

Cheers, Lars.

Hi Lars, thanks for your reply, I've attached my scene description.
Sorry but I do not have an ftp where to upload them although thwy are
very very small files...

To reply your questions:
I do not have any bright ring in the source image which is an angular
fisheye view (180 fov) generated in radiance of a CIE sunny sky using
the option -s (I do not expect this technique to be of any use for
direct sunlight, only for the diffuse component) the image is uncropped
and the cal file says: Projection of 2d image fisheye picture to source,
applies to angular fisheye like Nikon 180 deg.
Although I cannot find any geometrical distortion, it's only the value
nearby the horizon that goes up.
For point 2, I have the error of 1% nearby the centre (really where the
"sun" should be) and then it goes up as it approaches the horizon...

Thanks,

Giovanni

test00_sky.rad (279 Bytes)

test00.rif (494 Bytes)

test00.rad (109 Bytes)

···

-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Lars
O. Grobe
Sent: 08 September 2009 19:03
To: Radiance general discussion
Subject: Re: [Radiance-general] sky mapping

Hi!

It seems to work fairly well (the error at the centre is about 1% of
the original image) but

1) a bright circle forms along the horizon

2) the remapped image is constantly less bright than the original one

It would be nice if you could share the scene as well as the resulting
image by e.g. uploading it to a web oder ftp server. Please not by
posting any attachments.. If you have a bright ring in the result - do
you have a bright ring in your source image, e.g. at the edge of the
image circle? How did you prepare the image, do you crop it? Does the
cal-file expect a 180 fov. and does your image provide that?

Under 2) you write that the image is less bright then the original, but
at the centre you have only an error of 1% - do you have a constant
error of 1%?

Are you sure that the fisheye projection assumed by the script assumes
your camera/lens projection?

Does the source image include direct sun (visible)?

Cheers, Lars.

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

Hi Giovanni,

I have used fisheye.cal, which comes with radiance. I also used the
angular fisheye with the vta-view, as you did. The only important point
was to translate the mapping, as fisheye.cal expected the image center
to be +y (north in most models), while for sky capturing you point up.

The mapping for me is

void colorpict skyhdrmapping
11 red green blue sky.hdr fisheye.cal fish_u fish_v -rx -90 -rz 180
0
0

And sky.hdr has the north on the top when seen as a 2d image.

I am not sure whether Peter's cal-file does any more advanced stuff
related to his fisheye - it could be that he is correcting for some
vignetting stuff that does not exist in you synthetic image. That would
explain the higher pixel values close to the edge.

Cheers,

Lars.

Lars,
Thanks a lot for this,
It seems to work although in a slightly modified version (rx +90)
The results are much closer in the central part of the image (I assume
the difference is only because of interpolation or anti aliasing
effects...)And the brightest band at the horizon has almost disappeared
(although not entirely... again is the anti aliasing to blame?)
I've also noticed that for this process to work it is crucial that the
original sky image has a normalized exposure (1) otherwise all the
values are scaled down...

Thanks again,

Giovanni

···

-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Lars
O. Grobe
Sent: 09 September 2009 02:44
To: Radiance general discussion
Subject: RE: [Radiance-general] sky mapping

Hi Giovanni,

I have used fisheye.cal, which comes with radiance. I also used the
angular fisheye with the vta-view, as you did. The only important point
was to translate the mapping, as fisheye.cal expected the image center
to be +y (north in most models), while for sky capturing you point up.

The mapping for me is

void colorpict skyhdrmapping
11 red green blue sky.hdr fisheye.cal fish_u fish_v -rx -90 -rz 180
0
0

And sky.hdr has the north on the top when seen as a 2d image.

I am not sure whether Peter's cal-file does any more advanced stuff
related to his fisheye - it could be that he is correcting for some
vignetting stuff that does not exist in you synthetic image. That would
explain the higher pixel values close to the edge.

Cheers,

Lars.

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