Strip assembly panoramas with a digital slr -- for stereo etc

[panovrx]

http://www.mediavr.com/cerberusr.htm

This is a (unretouched) spherical pano of a cave but not stitched in
the usual way ie. blended images -- it is assembled in a few seconds
out of 120 (3 degree) very narrow vertical images strips with no
blending at all  -- using the mosaic tool of the excellent
Stereophoto Maker.

It is the right shot of a stereo pair. It is the middle exposure of
bracketed sequences. I havent enfused the other exposure panos with
it yet.
It was shot with a very accurate 120 step indexing head I made from a
large gear with a strip of brass clicking into the teeth

The camera here is to the right of the zero parallax point as it
rotates by about 7 cms but still because the steps are so small
perspective jumps are mostly invisible and the light (surprisingly
one might think) seems constant across the joins.

Each strip is a 3 by 180 degree equirectangular strip from a 5D/Nikkor
fisheye -- generated very quickly in PTGui or via a script with the
Panotools plugin.

Why would you want to make panoramas this way --

-- well you can totally automate the stitching process -- it is more
forgiving of slight positioning errors than standard template
stitching (here with stereo panoramas the mispositioning is extreme
compared with standard stitching practice  - yet still it stitches ok
automatically)
-- though you must be careful with the constancy of the alignment of
the camera tilt and roll with the rotation axis -- I use a digital
level to check)

.. it is good for stereo panoramas using either the two cameras (or
one camera with shift) or single camera/single rotation methods

.. it is good for scene contrast as you can put a custom lens hood on
the lens to give a narrow strip view of the scene ... and hence is
good for hdr panoramas too.

Peter Murphy
http://www.mediavr.com/blog

[Roger D. Williams]

On Mon, 28 Jul 2008 09:23:03 +0900, panovrx <[hidden email]> wrote:

> http://www.mediavr.com/cerberusr.htm
>
> This is a (unretouched) spherical pano of a cave but not stitched in
> the usual way ie. blended images -- it is assembled in a few seconds
> out of 120 (3 degree) very narrow vertical images strips with no
> blending at all  -- using the mosaic tool of the excellent
> Stereophoto Maker.

Very, very impressive. Both the quality and the assembly that takes
only a few seconds. And the mention of Stereophoto Maker is of great
interest to me, as stereo photos are (apart from panoramas) my
"other" great interest in photography.

> It is the right shot of a stereo pair. It is the middle exposure of
> bracketed sequences. I havent enfused the other exposure panos with
> it yet.
> It was shot with a very accurate 120 step indexing head I made from a
> large gear with a strip of brass clicking into the teeth

Oh dear. That sounds totally beyond me.

> The camera here is to the right of the zero parallax point as it
> rotates by about 7 cms but still because the steps are so small
> perspective jumps are mostly invisible and the light (surprisingly
> one might think) seems constant across the joins.

Yes, surprising indeed.

> Each strip is a 3 by 180 degree equirectangular strip from a 5D/Nikkor
> fisheye -- generated very quickly in PTGui or via a script with the
> Panotools plugin.

Hmmm. So as I have a Nikon 10.5mm fisheye I would need the new Nikon
700 (full frane) to take advantage of this method. "Oh dear" squared.

> Why would you want to make panoramas this way --
>
> -- well you can totally automate the stitching process -- it is more
> forgiving of slight positioning errors than standard template
> stitching (here with stereo panoramas the mispositioning is extreme
> compared with standard stitching practice  - yet still it stitches ok
> automatically)
> -- though you must be careful with the constancy of the alignment of
> the camera tilt and roll with the rotation axis -- I use a digital
> level to check)

Do you think the latest auto horizon indicators provided with Nikon
cameras (and maybe others, I only follow Nikon developments) would be
accurate enough?

> .. it is good for stereo panoramas using either the two cameras (or
> one camera with shift) or single camera/single rotation methods
>
> .. it is good for scene contrast as you can put a custom lens hood on
> the lens to give a narrow strip view of the scene ... and hence is
> good for hdr panoramas too.

Aha! That's another good point in its favor. Although don't you need
to shave OFF the hood to get the 180-degree FOV? Oh, I see, a "custome"
lens hood. Hmmm. Shave only two of the four petals?

Another advantage is that only a central strip of the image is used,
which makes best use of the region of highest image quality. Except
for the zenith and nadir, where it is much less important. I used to
get the same advantage from my Widelus swing-lens panorama camera,
which took a narrow slit from the middle of an image created by a
26mm lens. The image quality always astonished me.

Fascinating, Peter. Please keep us posted. If you have a mailing list
or any regular way of keeping people up to date and the news isn't
relevant to PanoToolsNG, perhaps you would kindly put me on the list?
I'd hate to miss out on this.

The only disadvantage I can think of is that it's not going to be
easy to go on a day's shoot and take, say, a dozen panoramas or so.
Or at least hot without a LOT of external memory.

Roger W.

--
Work: www.adex-japan.com
Play: www.usefilm.com/member/roger

[Kathy Wheeler]

On 28/07/2008, at 10:23 AM, panovrx wrote:

> http://www.mediavr.com/cerberusr.htm
>
> This is a (unretouched) spherical pano of a cave but not stitched in
> the usual way ie. blended images -- it is assembled in a few seconds
> out of 120 (3 degree) very narrow vertical images strips with no
> blending at all

That is beautiful and very interesting.

> It was shot with a very accurate 120 step indexing head I made from a
> large gear with a strip of brass clicking into the teeth

How large is "large"?

Please keep us posted.

Cheers,
KathyW.

[John Riley-2]

It is hard to believe that would work so well, but it looks great!  
Could you post some of the individual images and their remapped  
versions?

Thanks for sharing this,

John

John Riley
[hidden email]
[hidden email]




On Jul 27, 2008, at 8:23 PM, panovrx wrote:

> http://www.mediavr.com/cerberusr.htm
>
> This is a (unretouched) spherical pano of a cave but not stitched in
> the usual way ie. blended images -- it is assembled in a few seconds
> out of 120 (3 degree) very narrow vertical images strips with no
> blending at all -- using the mosaic tool of the excellent
> Stereophoto Maker.
>
> It is the right shot of a stereo pair. It is the middle exposure of
> bracketed sequences. I havent enfused the other exposure panos with
> it yet.
> It was shot with a very accurate 120 step indexing head I made from a
> large gear with a strip of brass clicking into the teeth
>
> The camera here is to the right of the zero parallax point as it
> rotates by about 7 cms but still because the steps are so small
> perspective jumps are mostly invisible and the light (surprisingly
> one might think) seems constant across the joins.
>
> Each strip is a 3 by 180 degree equirectangular strip from a 5D/Nikkor
> fisheye -- generated very quickly in PTGui or via a script with the
> Panotools plugin.
>
> Why would you want to make panoramas this way --
>
> -- well you can totally automate the stitching process -- it is more
> forgiving of slight positioning errors than standard template
> stitching (here with stereo panoramas the mispositioning is extreme
> compared with standard stitching practice - yet still it stitches ok
> automatically)
> -- though you must be careful with the constancy of the alignment of
> the camera tilt and roll with the rotation axis -- I use a digital
> level to check)
>
> .. it is good for stereo panoramas using either the two cameras (or
> one camera with shift) or single camera/single rotation methods
>
> .. it is good for scene contrast as you can put a custom lens hood on
> the lens to give a narrow strip view of the scene ... and hence is
> good for hdr panoramas too.
>
> Peter Murphy
> http://www.mediavr.com/blog
>
>
>

[Non-text portions of this message have been removed]

[PanoToolsNG.10.m8]

 

)-----Original Message-----
)From: Roger D. Williams
)Sent: Monday, 28 July 2008 10:22
)
)On Mon, 28 Jul 2008 09:23:03 +0900, panovrx <[hidden email]> wrote:
)
)> http://www.mediavr.com/cerberusr.htm
)>
)
)
)> Each strip is a 3 by 180 degree equirectangular strip from a
)5D/Nikkor
)> fisheye -- generated very quickly in PTGui or via a script with the
)> Panotools plugin.
)
)Hmmm. So as I have a Nikon 10.5mm fisheye I would need the new
)Nikon 700 (full frane) to take advantage of this method. "Oh
)dear" squared.
)
)
)
)> .. it is good for stereo panoramas using either the two cameras (or
)> one camera with shift) or single camera/single rotation methods
)>
)> .. it is good for scene contrast as you can put a custom
)lens hood on
)> the lens to give a narrow strip view of the scene ... and hence is
)> good for hdr panoramas too.
)
)Aha! That's another good point in its favor. Although don't
)you need to shave OFF the hood to get the 180-degree FOV? Oh,
)I see, a "custome"
)lens hood. Hmmm. Shave only two of the four petals?

If I recall correctly, you still get 180 degrees FOV across the diagonal
without shaving the hood. As you are only using about 3 degrees of ea.
image, you could mount the camera with a roll, such that the diagonal across
the frame is aligned vertically.

[Roger D. Williams]

On Mon, 28 Jul 2008 10:29:52 +0900, <[hidden email]>  
wrote:

> )> .. it is good for scene contrast as you can put a custom
> )lens hood on
> )> the lens to give a narrow strip view of the scene ... and hence is
> )> good for hdr panoramas too.
> )
> )Aha! That's another good point in its favor. Although don't
> )you need to shave OFF the hood to get the 180-degree FOV? Oh,
> )I see, a "custome"
> )lens hood. Hmmm. Shave only two of the four petals?
>
> If I recall correctly, you still get 180 degrees FOV across the diagonal
> without shaving the hood. As you are only using about 3 degrees of ea.
> image, you could mount the camera with a roll, such that the diagonal  
> across
> the frame is aligned vertically.

Yes, that's an excellent point. Of course it still assumes a full-frame
image. The expense involved in that purchase is quite a consideration.

Roger W.

--
Work: www.adex-japan.com
Play: www.usefilm.com/member/roger

[Milko K. Amorth]

Roger,
> Yes, that's an excellent point. Of course it still assumes a full-frame
> image.
The stock 10.5 on a 1.5 crop sensor (Fuji,Minolta,Nikon, Pentax) yields
180° fov diagonally. Anything smaller than that yields less.

Cheers, Milko



--
Milko K.Amorth
ph:604.561.5101
fx:604.909.5125

www.VRCanada.ca
360° Immersive Imaging
Skype: VRdundee

[PanoToolsNG.10.m8]

 

)-----Original Message-----
)From: Roger D. Williams
)Sent: Monday, 28 July 2008 12:05
)
)> )> .. it is good for scene contrast as you can put a custom
)> )lens hood on
)> )> the lens to give a narrow strip view of the scene ... and hence is
)> )> good for hdr panoramas too.
)> )
)> )Aha! That's another good point in its favor. Although don't
)> )you need to shave OFF the hood to get the 180-degree FOV? Oh,
)> )I see, a "custome"
)> )lens hood. Hmmm. Shave only two of the four petals?
)>
)> If I recall correctly, you still get 180 degrees FOV across
)the diagonal
)> without shaving the hood. As you are only using about 3
)degrees of ea.
)> image, you could mount the camera with a roll, such that the
)diagonal  
)> across
)> the frame is aligned vertically.
)
)Yes, that's an excellent point. Of course it still assumes a full-frame
)image. The expense involved in that purchase is quite a consideration.
)

I thought the 180deg diagonal was with the smaller sensor size?

Btw: there used to be a program called snapdv that was able to make
panoramas from a camcorder. You would pan slowly around and I believe the
program would take slivers out of each of the frames of the avi to create
the panorama.  

Cheers,
Darren.

[Roger D. Williams]

On Mon, 28 Jul 2008 12:18:15 +0900, Milko K. Amorth  
<[hidden email]> wrote:

> Roger,
>> Yes, that's an excellent point. Of course it still assumes a full-frame
>> image.
> The stock 10.5 on a 1.5 crop sensor (Fuji,Minolta,Nikon, Pentax) yields
> 180° fov diagonally. Anything smaller than that yields less.
>
> Cheers, Milko

Thanks, Milko, that's just what I wanted to hear. Looks like I can manage  
with
a D300!

Roger



--
Work: www.adex-japan.com
Play: www.usefilm.com/member/roger

[panovrx]

Here are some pics of the rig I shot this pano with:
http://www.mediavr.com/stereohead2.jpg
this is the whole assembly

http://www.mediavr.com/stereohead3.jpg
this is the indexing mechanism

http://www.mediavr.com/stereohead1.jpg
this is the miniature gearhead motor that I use to drive
the gear from click to click (I control it with a remote button)
-- I listen for the 2nd click to know when to stop pressing
-- with the little drive gear I filed some teeth off to handle
the momentum of the gear head motor after the current is off
so it doesnt press the main gear


there are 240 teeth = 1.5degrees
normally I move the gear on two clicks at a time = 3 degrees

Peter Murphy
http://www.mediavr.com/blog



>
> http://www.mediavr.com/cerberusr.htm
>
> This is a (unretouched) spherical pano of a cave but not stitched
in
> the usual way ie. blended images -- it is assembled in a few
seconds
> out of 120 (3 degree) very narrow vertical images strips with no
> blending at all  -- using the mosaic tool of the excellent
> Stereophoto Maker.
>
> It is the right shot of a stereo pair. It is the middle exposure of
> bracketed sequences. I havent enfused the other exposure panos with
> it yet.
> It was shot with a very accurate 120 step indexing head I made from
a
> large gear with a strip of brass clicking into the teeth
>
> The camera here is to the right of the zero parallax point as it
> rotates by about 7 cms but still because the steps are so small
> perspective jumps are mostly invisible and the light (surprisingly
> one might think) seems constant across the joins.
>
> Each strip is a 3 by 180 degree equirectangular strip from a
5D/Nikkor
> fisheye -- generated very quickly in PTGui or via a script with the
> Panotools plugin.
>
> Why would you want to make panoramas this way --
>
> -- well you can totally automate the stitching process -- it is
more
> forgiving of slight positioning errors than standard template
> stitching (here with stereo panoramas the mispositioning is extreme
> compared with standard stitching practice  - yet still it stitches
ok
> automatically)
> -- though you must be careful with the constancy of the alignment
of
> the camera tilt and roll with the rotation axis -- I use a digital
> level to check)
>
> .. it is good for stereo panoramas using either the two cameras (or
> one camera with shift) or single camera/single rotation methods
>
> .. it is good for scene contrast as you can put a custom lens hood
on
> the lens to give a narrow strip view of the scene ... and hence is
> good for hdr panoramas too.
>
> Peter Murphy
> http://www.mediavr.com/blog
>

[Fabio Bustamante-2]

Wouldn't this be the closest digital version of those rotating panoramic
film cameras that used to slide the film as they turned around?

Impressive result indeed!

panovrx wrote:
> Each strip is a 3 by 180 degree equirectangular strip from a 5D/Nikkor
> fisheye -- generated very quickly in PTGui or via a script with the
> Panotools plugin.
>
>
>  

[bohonus]

--- In [hidden email], Fabio Bustamante <contato@...> wrote:
>
> Wouldn't this be the closest digital version of those rotating panoramic
> film cameras that used to slide the film as they turned around?
>
> Impressive result indeed!

Not quite, though it certainly borrows from the concept. The closest digital version of the
cameras that rotate using film to capture a panoramic image is a camera using a trilinear
CCD that rotates electronically capturing a single vertical row of pixels successively until an
entire image is created. In the case of the Panoscan- up to 65,000 discrete steps/exposures.
Try finding a toothed gear with THAT many teeth, or clicking the shutter that many times
with a DSLR :)

See the following sites for examples of this type of digital panoramic camera technology.

http://www.panoscan.com/
http://www.spheron.com/
http://www.roundshot.ch/

[Flemming V. Larsen]

Interesting idea which remind me of one of my first pano-setup with a
video-camera spinning (around 40 sec for 360 degree) on a home made
"motorhead" made from daughters old toy musicbox and stitched with the
Jutvision Software that took a slices from each frame of the videoseqence
and stitched it to a panorama.
Here is a sample: http://vtour.dk/vtb/8000/dgby/03.jpg
The resolution and sharpness wasn't that good - but as you can see the big
problem was repeating refleksions ( in one of the windows) .Could that also
be a prblem with your setup?


> Roger W:
> The only disadvantage I can think of is that it's not going to be
> easy to go on a day's shoot and take, say, a dozen panoramas or so.
> Or at least hot without a LOT of external memory.
>

If you use a lenscap with a naarow cut slit that only allows a 5-6 degree
strip of light I think with most of the picture being black the filesize
will be heavily reduces (in raw(nef) and jpeg).



Flemming

[crane-2]

Quoting "Roger D. Williams" <[hidden email]>:

> On Mon, 28 Jul 2008 09:23:03 +0900, panovrx <[hidden email]> wrote:
>
> Another advantage is that only a central strip of the image is used,
> which makes best use of the region of highest image quality. Except
> for the zenith and nadir, where it is much less important.

I always thought an elliptical or rectangular crop on the fisheye image in
ptgui would be handy for that reason.
maybe it is there? I still have version 6.0.3

regards

mick


----------------------------------------------
This mail sent through http://www.ukonline.net

[Erik Krause]

crane-2 wrote:

I always thought an elliptical or rectangular crop on the fisheye image in
ptgui would be handy for that reason.
maybe it is there? I still have version 6.0.3

You can use fullframe for circular fisheyes in order to crop rectangular. Could be it works only if the images are loaded as landscape, since the FoV refers to the horizontal width in PTGui.

Perhaps hugin has an advantage there: It doesn't refer the FoV to the crop area since it uses "Selection" cropping. See http://wiki.panotools.org/Panotools_internals#Cropping for details

best regards

[Roger D. Williams]

Patrick,

How essential to your concept of the multiple-strip image route to  
360-degree
panoramas is the extreme narrowness of the strips?

How did you come upon the 120-step gear and its 3-degree FOV?  Was this the
result of unsatisfactory work on fewer, wider strips?

I am wondering if there might be a "sweet spot" somewhere between 3 degrees
and the 60 to 90 degrees used for conventional stitching. Intuitively, this
would have to be nearer to 3 degrees than 60 or 90, but I can't help
thinking that fewer strips--if they could be made to work as well--would be
better.

My guess is that there are trade offs between the number of images and the
narrowness of the strips and the ease of stitching them. And I wondered if
you had looked into that.

Roger W.


--
Work: www.adex-japan.com
Play: www.usefilm.com/member/roger

[panovrx]

Roger

I developed this kind of gear for stereo panoramas so I only know
about requirements for that with this strip approach. viz. 3 degrees
is the maximum if the the closest object is say 1m. Any closer and
you would need more shots. If the closest thing was 2m away one could
use maybe 6 degrees. I chose the gear I did because it was the one I
had (from an old precision potentiometer I think)  and its number of
teeth divided nicely into 360 .

With no parallax rotation you could use fewer steps but other
discontinuities than position parallax effects would emerge -
calibration errors would be more apparent and light jumps (from flare
differences etc) more likely maybe.

Peter Murphy
www.mediavr.com/blog

--- In [hidden email], "Roger D. Williams" <roger@...>
wrote:
>
> Patrick,
>
> How essential to your concept of the multiple-strip image route to  
> 360-degree
> panoramas is the extreme narrowness of the strips?
>
> How did you come upon the 120-step gear and its 3-degree FOV?  Was
this the
> result of unsatisfactory work on fewer, wider strips?
>
> I am wondering if there might be a "sweet spot" somewhere between 3
degrees
> and the 60 to 90 degrees used for conventional stitching.
Intuitively, this
> would have to be nearer to 3 degrees than 60 or 90, but I can't help
> thinking that fewer strips--if they could be made to work as well--
would be
> better.
>
> My guess is that there are trade offs between the number of images
and the
> narrowness of the strips and the ease of stitching them. And I
wondered if
> you had looked into that.
>
> Roger W.
>
>
> --
> Work: www.adex-japan.com
> Play: www.usefilm.com/member/roger
>

[panovrx]

Re stereo panorama viewing -- http://www.gali-
3d.com/archive/articles/StereoPanorama/StereoPanorama.php is a cool
stereo pano viewer that I often use for checking retouching of stereo
pano pairs and I got this working tonight with the Iz3d display.
http://www.iz3d.com -- after installing - with much fiddling - the
new Iz3d Opengl stereo driver

The Iz3d screen is the clearest I have seen stereo panos on a 3d
display. If Iz3d could only reduce the ghosting a bit with the screen
they would have killer product imo.

Peter Murphy





--- In [hidden email], "panovrx" <mediavr@...> wrote:
>
> http://www.mediavr.com/cerberusr.htm
>
> This is a (unretouched) spherical pano of a cave but not stitched
in
> the usual way ie. blended images -- it is assembled in a few
seconds
> out of 120 (3 degree) very narrow vertical images strips with no
> blending at all  -- using the mosaic tool of the excellent
> Stereophoto Maker.
>
> It is the right shot of a stereo pair. It is the middle exposure of
> bracketed sequences. I havent enfused the other exposure panos with
> it yet.
> It was shot with a very accurate 120 step indexing head I made from
a
> large gear with a strip of brass clicking into the teeth
>
> The camera here is to the right of the zero parallax point as it
> rotates by about 7 cms but still because the steps are so small
> perspective jumps are mostly invisible and the light (surprisingly
> one might think) seems constant across the joins.
>
> Each strip is a 3 by 180 degree equirectangular strip from a
5D/Nikkor
> fisheye -- generated very quickly in PTGui or via a script with the
> Panotools plugin.
>
> Why would you want to make panoramas this way --
>
> -- well you can totally automate the stitching process -- it is
more
> forgiving of slight positioning errors than standard template
> stitching (here with stereo panoramas the mispositioning is extreme
> compared with standard stitching practice  - yet still it stitches
ok
> automatically)
> -- though you must be careful with the constancy of the alignment
of
> the camera tilt and roll with the rotation axis -- I use a digital
> level to check)
>
> .. it is good for stereo panoramas using either the two cameras (or
> one camera with shift) or single camera/single rotation methods
>
> .. it is good for scene contrast as you can put a custom lens hood
on
> the lens to give a narrow strip view of the scene ... and hence is
> good for hdr panoramas too.
>
> Peter Murphy
> http://www.mediavr.com/blog
>

[Roger D. Williams]

Apologies for addressing Patrick when it should have been Peter...

On Tue, 29 Jul 2008 20:04:40 +0900, panovrx <[hidden email]> wrote:

> Roger
>
> I developed this kind of gear for stereo panoramas so I only know
> about requirements for that with this strip approach. viz. 3 degrees
> is the maximum if the the closest object is say 1m. Any closer and
> you would need more shots. If the closest thing was 2m away one could
> use maybe 6 degrees. I chose the gear I did because it was the one I
> had (from an old precision potentiometer I think)  and its number of
> teeth divided nicely into 360 .

I thought that might be the case. Still, a reduction in the number of
shots from 120 to 60 sounds good to me. On the other hand, if I could
get into the way of shooting the full 120 I would know I could take
stereo pairs, something I have never been able to do successfully with
panoramas (well, except for those taken with my Fuji X-1 and 45mm lens
using a slide bar). But these, though nice, are very different from a
full 360 immersive panorama. <sigh>

> With no parallax rotation you could use fewer steps but other
> discontinuities than position parallax effects would emerge -
> calibration errors would be more apparent and light jumps (from flare
> differences etc) more likely maybe.

It's all about trade-offs, isn't it! Like so much else in life...

Roger W.


--
Work: www.adex-japan.com
Play: www.usefilm.com/member/roger

[Roger D. Williams]

On Wed, 30 Jul 2008 00:04:24 +0900, panovrx <[hidden email]> wrote:

> Re stereo panorama viewing -- http://www.gali-
> 3d.com/archive/articles/StereoPanorama/StereoPanorama.php is a cool
> stereo pano viewer that I often use for checking retouching of stereo
> pano pairs and I got this working tonight with the Iz3d display.
> http://www.iz3d.com -- after installing - with much fiddling - the
> new Iz3d Opengl stereo driver
>
> The Iz3d screen is the clearest I have seen stereo panos on a 3d
> display. If Iz3d could only reduce the ghosting a bit with the screen
> they would have killer product imo.

Peter, it doesn't look as if the iZ3D display is available in Japan,
although I expect that with gaming fever at the high temperature it
is here there must be some local equivalent. I'll start looking. I
have kept a couple of 19" CRT monitors specifically so that I could
use their fast switching speed for "gated" viewing--and LCDs were
not capable of high enough speeds when I checked a few years ago.
But they are such space HOGS I would like to get rid of them. Two
19" CRT displays on a desk don't leave much room for anything else.

People who have never seen a 3D panorama have not seen panoramas at
their best. If you are "hooked" on immersive panoramas and haven't
seen them in 3D you owe it to yourself to do so. There now, I hope
that puts me back on topic. <grin>

Roger W.

--
Work: www.adex-japan.com
Play: www.usefilm.com/member/roger