PanoTools mailing list archive
|Sender:||Bjørn K Nilssen|
|Date/Time:||Thu, 06 Jul 2006 22:21:24 +0200|
|Subject:||Re: Pano with 3D model|
On 6 Jul 2006 at 19:22, Peter Nyfeler wrote:
> Hi Björn
Thanks for all the nice replies. It was a bit surprising though, as I thought that some
of you gurus had done similar things before?
> It is exactly what like to and ask some time ago.
> Ohh....and I forgot to thank you for the detailed answer, sorry!
> How long does it take for you as an experienced 3D designer to make
> such a panorama?
In this case it was an old project/model that my client wanted more images from. As I had
started making photographic panoramas in the mean time I suggested to him that I should
also make a pano with 3D model this time, to show the nice view.
I guess it took a couple of days to make the 3D model, and then I had to add some
interiors and furniture because the camera would be so close to the windows - a couple of
hours extra? Texturing and setting lights took a couple of hours too. Rendering,
stitching and compositing in PS maybe an additional 6h? It is a 40min drive each way to
the site, so let's say 4 days all in all?
> And would you be so kind and make a short description how you made it?
It is actually quite easy.
First I used a pole so that the camera was at eyeheight for a person standing at the
balcony, about 5m+ up from the ground, and I was standing approximately right below the
balcony. 4 shots around and stitching in PTgui.
In Lightwave (LW) I made a quite simple, large sphere and applied the pano on the inside
with spherical mapping (had to flip it to get it right).
I set the surface to high Luminosity, and 0 on everything else, and set the object to
neither cast nor receive any shadows.
Then I moved the sphere so that its center was exactly on the same spot as the camera in
LW. I used parenting to do it automatically.
The sphere was set to be unseen by camera, but visible to rays. This was done so that the
glass should reflect the pano, but not be visible in the render so that it would
hide/ruin the alpha channel, which I needed for compositing later.
The render was done with radiosity (FPrime), and the sphere/pano actually adds to the
lighting of the scene (HDR-lite ;) softening the shadows and tinting the colors slightly.
With a real HDR pano it would be enough to use only the pano as light source, because the
sun would be strong enough to cast shadows into the scene, and the other colors would not
be strong enough to color the model green (from the grass) etc, which is what would
happen if I upped the luminosity of the pano enough to light the scene with the HDR-lite
version alone. I guess I could have made a simple HDR version with a strong sun and
desaturated colors for lighting the scene?
The camera was set to have a fov of 110 degrees, 2048x2048pixels, and it was keyframed
with 90 degrees yaw between each of the first 4 frames, and then a nadir and zenith.
Stitches very well in PTgui with no need for optimizing anything ;)
PTgui didn't like the alpha channel of the png32 (or was it tga32 I used?) files though,
so I had to stitch with 24 bit versions, and then again for the alpha channel.
After rendering and stitching the model + a separate mask from the alpha channel I loaded
all 3 files into PS.
The pano as background,
then the render, with the alpha applied as layer mask
and then a copy of the pano on top, with layer mask to reveal the model.
Using 3 layers makes it very much easier if I need to rerender the model, because I
already have the needed mask in the top layer mask, instead of in the rendered layer.
I had to do some horizontal offsetting to align the pano with the renders though.
I can render a full spherical pano directly in LW, but then I don't get any alpha
channel. That's why I used the cubic stitching approach.
That's all :)
#removed# // Bjørn Kåre Nilssen http://bknilssen.no/
Yahoo! Groups Links
<*> To visit your group on the web, go to:
<*> To unsubscribe from this group, send an email to:
<*> Your use of Yahoo! Groups is subject to: