0.61 update of Panini perspective tool available

[Tom Sharpless]

I've posted updated source code and Win32 binary for the Panini
perspective tool at http://sourceforge.net/projects/pvqt/. A Mac
binary should be up in a day or 2 when Harry is feeling better.

This update is version 0.61.77. It fixes a couple of bugs and
improves the GUI in several ways (including a nice logo thanks to
Harry). The only new functionality is a menu item that redisplays the
wire frame models instead of an image.

Enjoy, Tom

[Bart.van.Andel]

Hi Tom,

Thanks for bringing regular updates! Unfortunately, I still have
issues displaying the example cubic panoramas. Loading from JPGs and
QTVR gives different errors on screen though. It might be the OpenGL
implementation on my machine, although I doubt that, because I've
never seen such errors in other applications using OpenGL. Here's an
image showing 4 screenshots: both QTVR and cube faces, both with
initial view and super fish at 260 degrees FoV.

http://lh3.ggpht.com/_Fel5i-Q8T8g/SWXolMtRS5I/AAAAAAAABxA/i78G0TsTEw4/2009.01.08%20Panini%20cubic%20pano%20errors.jpg

As you can see, the same face image is shown on every cube face,
although with slightly different displaying errors. I assume you are
still using OpenGL's builtin solution for the cube thing? I guess this
one causes the trouble.

I'll see if I can figure out more details later.

Good luck,
Bart

[Tom Sharpless]

Thanks, Bart

On Jan 8, 6:57 am, "Bart.van.Andel" <bavanan...@gmail.com> wrote:
>
> http://lh3.ggpht.com/_Fel5i-Q8T8g/SWXolMtRS5I/AAAAAAAABxA/i78G0TsTEw4...

Aaaarghh! Mac users are reporting all sorts of problems with cubic
format, you are not alone...

>
> As you can see, the same face image is shown on every cube face,
> although with slightly different displaying errors. I assume you are
> still using OpenGL's builtin solution for the cube thing?

Yes. It has been a standard feature since OpenGL 1.4, and always works
fine on my computers, under Windows and Linux. And I wouldn't have
any idea how to set up mapping of multiple texture images onto one
surface. If you do know, please tell me.

> I guess this one causes the trouble.

But why only on Macs? And with 3 different video systems at least?

Am I safe in assuming that you can display the non-cubic formats OK?

>
> I'll see if I can figure out more details later.

Any help you can give would be most welcome. I am not a Mac developer
and have no Mac to test on. At this point I have no idea what is
wrong. Some other problems that have been reported:
-- defective areas shaped like parts of the edges of the cube
-- failed cubic loads with error messages and/or black display
-- cancelling a cubic load at the fileselector makes subsequent loads
fail (this has got to be my fault, but it only has been seen on a
Mac).

Regards, Tom

[Bart.van.Andel]

> Aaaarghh!  Mac users are reporting all sorts of problems with cubic
> format, you are not alone...

> But why only on Macs?  And with 3 different video systems at least?

For the record, I don't have a Mac (I only wish I had one, haha). I'm
running WinXP SP3 and just updated my graphic card drivers (ATI Radeon
9600SE) to the newest version. I just installed the drivers, not that
piece of memory consuming bloatware called Catalyst.

> I wouldn't have
> any idea how to set up mapping of multiple texture images onto one
> surface. If you do know, please tell me.

Well, basically they're just 6 rectangular projections 90 degrees
apart, which shouldn't be too difficult to backproject onto the
panosphere. I guess you're already doing the same thing when loading a
rectangular image?

> Am I safe in assuming that you can display the non-cubic formats OK?

Correct. Only trouble with cubics.

> > I'll see if I can figure out more details later.
>
> Any help you can give would be most welcome.  I am not a Mac developer
> and have no Mac to test on.

Me neither, unfortunately. I've read through the source code a bit and
couldn't find anything wrong there. Moreover, I compared your code to
some of the code available on other websites (by googling), and if not
pretty much an exact copy, it's still fairly the same amount and order
of operations. Strange.

If you look closer at the picture I linked to in my previous post,
it's not only that the same image is shown on every face. It's shown
with a *different* error on every face. The images are shifted in both
X and Y directions (in texture coordinates), and have border issues
which differ from face to face. It popped into my mind that something
might be wrong with texture coordinate generation internally, but this
does not quite explain the cloning behavior.

By the way this is the first time I dive into OpenGL, so don't expect
me doing wonders just yet ;)

Cheers,
Bart

[Peter Gawthrop]

Hi Tom,

0.61.79M gives me a "black hole" at the zenith and nadir when
displaying equirectangulars. The equivalent QTVR is OK.

Best wishes

Peter.

> ______________________________________________
> This email has been scanned by Netintelligence
> http://www.netintelligence.com/email
>

[Tom Sharpless]

Hi Peter

On Jan 18, 9:47 am, Peter Gawthrop <pe...@gawthrop.net> wrote:
> Hi Tom,
>
>   0.61.79M gives me a "black hole" at the zenith and nadir when
>   displaying equirectangulars. The equivalent QTVR is OK.
>

:-P That's not a bug; it's a feature!

Really. The 0.61 update defaults to the panocylinder for non-cubic
images, so that you won't have to switch to make Pannini views. So
if you want to view full spherical images, you now have to switch back
to the panosphere. (Well, nothing is perfect.) Future releases may
do yaw and pitch before painting the panosurface, which would let you
see zenith and nadir (when available) on the cylinder too.

I've been experimenting with Mathmap, trying to replicate Panini's
projections (using your "erect_erect.mm" to set the view direction;
thanks) and it is stimulating several ideas about better generalized
Pannini projections. Command line Mathmap looks like an ideal "back
end" to render Panini views at source resolution. Or maybe Panini
could become a "front end" for Mathmap, which is clearly the more
capable program. My goal is an interactive system that can "tweak"
the basic Pannini projection to render a given view perfectly -- just
as the vedutisti did.

Regards, Tom