Quantcast

Hyperstereo pole panorama -- anaglyph

classic Classic list List threaded Threaded
29 messages Options
12
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Hyperstereo pole panorama -- anaglyph

panovrx-2
http://www.mediavr.com/artgallerystereoana.htm

from about 4.5m with 50 shots in 18 seconds on a rotating twin 5D/10.5mm rig with camera separation 25cm
...like
http://www.youtube.com/watch?v=EwzCMPeS6JE
except with the cameras further apart and hanging down from a bar and rotating slower

When you are far enough away from the nearest feature, like high on a pole, quite large disparity pano pairs can be stitched

PeterM

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

Willy Kaemena
wow  that looks unbelievable deep !!!   wonderful


Willy


On Dec 19, 2010, at 0:35, panovrx wrote:

> http://www.mediavr.com/artgallerystereoana.htm
>
> from about 4.5m with 50 shots in 18 seconds on a rotating twin 5D/10.5mm rig with camera separation 25cm
> ...like
> http://www.youtube.com/watch?v=EwzCMPeS6JE
> except with the cameras further apart and hanging down from a bar and rotating slower
>
> When you are far enough away from the nearest feature, like high on a pole, quite large disparity pano pairs can be stitched
>
> PeterM
>
>



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

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

Ned Chiariello
In reply to this post by panovrx-2
This is the best 3D panorama yet, I'm hooked!

On Dec 18, 2010, at 4:35 PM, panovrx wrote:

> http://www.mediavr.com/artgallerystereoana.htm
>
> from about 4.5m with 50 shots in 18 seconds on a rotating twin 5D/10.5mm rig with camera separation 25cm
> ...like
> http://www.youtube.com/watch?v=EwzCMPeS6JE
> except with the cameras further apart and hanging down from a bar and rotating slower
>
> When you are far enough away from the nearest feature, like high on a pole, quite large disparity pano pairs can be stitched
>
> PeterM
>
>



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



------------------------------------

--
<*> Wiki: http://wiki.panotools.org
<*> User Guidelines: http://wiki.panotools.org/User_Guidelines
<*> Nabble (Web) http://n4.nabble.com/PanoToolsNG-f586017.html
<*> NG Member Map http://www.panomaps.com/ng
<*> Moderators/List Admins: [hidden email]
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/PanoToolsNG/

<*> Your email settings:
    Individual Email | Traditional

<*> To change settings online go to:
    http://groups.yahoo.com/group/PanoToolsNG/join
    (Yahoo! ID required)

<*> To change settings via email:
    [hidden email]
    [hidden email]

<*> To unsubscribe from this group, send an email to:
    [hidden email]

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

jrgen_schrader
In reply to this post by panovrx-2
A brilliant piece of frozen reality.
I like that woman taking the parking ticket from the vending machine and all the people walking around.

Jürgen

--- In [hidden email], "panovrx" <panovrx@...> wrote:

>
> http://www.mediavr.com/artgallerystereoana.htm
>
> from about 4.5m with 50 shots in 18 seconds on a rotating twin 5D/10.5mm rig with camera separation 25cm
> ...like
> http://www.youtube.com/watch?v=EwzCMPeS6JE
> except with the cameras further apart and hanging down from a bar and rotating slower
>
> When you are far enough away from the nearest feature, like high on a pole, quite large disparity pano pairs can be stitched
>
> PeterM
>


Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

Trausti Hraunfjord
A lot of depth all around.  Really good! ..... but the one thing I still
miss in those panos, is that the mouse cursor would drop and rise according
to the depth of the image... but I really don´t know how that could be
achieved.

Trausti


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

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

Wim Koornneef
In reply to this post by panovrx-2
Hello Peter,

It can be very difficult to determine the proper camera seperation on forehand and only on the location itself it is possible to judge the proper setting and even then it is not sure how it will work out.
In this case the 25 cm you choose is working fine, the depth is maximized for the scene, any more then 25 cm would apply to much 3D depth on the foreground (poles etc.) and any less then 25 cm would make the background (trees, buildings) to flat.
Apart from all technical details that are involved I think setting the proper camera seperation is perhaps one of the most difficult things.

The panorama itself shows a nice scene and there is no sign of ghosting at all.
Well done !

Wim
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

Wim Koornneef
PS,

When looking up in the pano and then rotating it seems that zenith is wobbling a lot, this could be caused by the parallax but it could be also that you limited the player.
I can also see the vFOV of the pano player is not symmetric, did you use an up tilt for the cameras or did you limit the down view ?

Wim
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

afrenyo
In reply to this post by panovrx-2
Super.
The height, the spot, the separation - all has worked pretty perfect here.

How many images did you use to stitch the panos.  Can I assume not all and the same for both?

Thanks.

On Dec 19, 2010, at 8:38 AM, [hidden email] wrote:

> http://www.mediavr.com/artgallerystereoana.htm



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

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

panovrx-2



--- In [hidden email], Andras Frenyo <andras@...> wrote:

>
> Super.
> The height, the spot, the separation - all has worked pretty perfect here.
>
> How many images did you use to stitch the panos.  Can I assume not all and the same for both?
>
> Thanks.
>
> On Dec 19, 2010, at 8:38 AM, [hidden email] wrote:
>
> > http://www.mediavr.com/artgallerystereoana.htm
>
>
It was a rather convoluted workflow but this method seemed to work well with this subject.
First I trimmed the L and R sequences, head and tail, to enough matched pairs to cover 360 and a little bit more.
About 56 images for each side. Then I check the sync by looking at action in the pairs. Then I loaded the L shots
and masked them in PTGui 9 so that only an area above the horizon and about 60 degrees wide was being used
for point finding. Then I calibrated the L shots to see if I could get a good solution. When it looked ok I levelled it.
Then I loaded the first of the R hand shots as well. I calibrated its yaw, pitch and roll against the first of the L shots.
Then I loaded the rest of the R shots.

Now I know what roll, pitch and yaw of the first L shot will be level. What fov and a,b, c of the fisheye lens will give
a good stitch, and what roll, pitch and yaw of the first R shot will make it parallized to the first L shot.

So now I set all the L shots to the same roll,pitch and yaw as the first L  shot and all the R shots to the same as
the first  R shot. So now all the image pairs are parallelized (like stereo pairs with the same roll and tilt and aligned
at the same infinity points),

Then I convert all the fisheye images to 20 degree wide, 180 degree high equi strips using these settings.

Then I load the equi L strips and optimize them (after masking the foreground for point finding)
 and stitch the final L panorama with Smartblend blending.
I output individual 360 panoramas for each L strip also.

Then I make a template of this solution and apply it to the R strips and output the finished R panorama and individual
panoramas for each R strip.

Smartblend will have blended differently in different action parts of the scene so you now have to fix those
(in a couple of places) with pasted bits from the individual  strip panoramas you have outputted.

Then I load the finished L and R stitched and fixed panoramas into PTGui and level them again each other and finetune
the scene levelling.



You will think this has too many steps and I could have stitched directly from the fisheyes  but I think the solution I got
with the second generation images (equi strips from the parallelised calibrated fisheyes) was quite a bit more accurate
than what I got directly from stitching the first L fisheye test stitch. But sometimes directly from the fisheyes is better.

Peter M


Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

Roger D Williams
On Tue, 21 Dec 2010 19:09:03 +0900, panovrx <[hidden email]> wrote:

>
>
>
> --- In [hidden email], Andras Frenyo <andras@...> wrote:
>>
>> Super.
>> The height, the spot, the separation - all has worked pretty perfect  
>> here.
>>
>> How many images did you use to stitch the panos.  Can I assume not all  
>> and the same for both?
>>
>> Thanks.
>>
>> On Dec 19, 2010, at 8:38 AM, [hidden email] wrote:
>>
>> > http://www.mediavr.com/artgallerystereoana.htm
>>
>>
> It was a rather convoluted workflow but this method seemed to work well  
> with this subject.

I found this both fascinating and extremely depressing. The result is
great, but that it took so much expertise and detailed work means that
I won't be able to produce anything nearly as good in this lifetime.
I can't imagine that Sony will ever automate anything like THIS!

Roger W.

--
Business: www.adex-japan.com
Pleasure: www.usefilm.com/member/roger
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

jrgen_schrader
In reply to this post by panovrx-2
Thank you very much, Peter, for the detailed insight.
Much appreciated.

After I had made some more tests with Wim's single camera solution I wondered how it would be possible to get the panoramas from to cameras with an offset leveled and parallelized so that they would make a useable stereo pair. Your description helped a lot.

Jürgen



--- In [hidden email], "panovrx" <panovrx@...> wrote:

>
>
>
>
> --- In [hidden email], Andras Frenyo <andras@> wrote:
> >
> > Super.
> > The height, the spot, the separation - all has worked pretty perfect here.
> >
> > How many images did you use to stitch the panos.  Can I assume not all and the same for both?
> >
> > Thanks.
> >
> > On Dec 19, 2010, at 8:38 AM, [hidden email] wrote:
> >
> > > http://www.mediavr.com/artgallerystereoana.htm
> >
> >
> It was a rather convoluted workflow but this method seemed to work well with this subject.
> First I trimmed the L and R sequences, head and tail, to enough matched pairs to cover 360 and a little bit more.
> About 56 images for each side. Then I check the sync by looking at action in the pairs. Then I loaded the L shots
> and masked them in PTGui 9 so that only an area above the horizon and about 60 degrees wide was being used
> for point finding. Then I calibrated the L shots to see if I could get a good solution. When it looked ok I levelled it.
> Then I loaded the first of the R hand shots as well. I calibrated its yaw, pitch and roll against the first of the L shots.
> Then I loaded the rest of the R shots.
>
> Now I know what roll, pitch and yaw of the first L shot will be level. What fov and a,b, c of the fisheye lens will give
> a good stitch, and what roll, pitch and yaw of the first R shot will make it parallized to the first L shot.
>
> So now I set all the L shots to the same roll,pitch and yaw as the first L  shot and all the R shots to the same as
> the first  R shot. So now all the image pairs are parallelized (like stereo pairs with the same roll and tilt and aligned
> at the same infinity points),
>
> Then I convert all the fisheye images to 20 degree wide, 180 degree high equi strips using these settings.
>
> Then I load the equi L strips and optimize them (after masking the foreground for point finding)
>  and stitch the final L panorama with Smartblend blending.
> I output individual 360 panoramas for each L strip also.
>
> Then I make a template of this solution and apply it to the R strips and output the finished R panorama and individual
> panoramas for each R strip.
>
> Smartblend will have blended differently in different action parts of the scene so you now have to fix those
> (in a couple of places) with pasted bits from the individual  strip panoramas you have outputted.
>
> Then I load the finished L and R stitched and fixed panoramas into PTGui and level them again each other and finetune
> the scene levelling.
>
>
>
> You will think this has too many steps and I could have stitched directly from the fisheyes  but I think the solution I got
> with the second generation images (equi strips from the parallelised calibrated fisheyes) was quite a bit more accurate
> than what I got directly from stitching the first L fisheye test stitch. But sometimes directly from the fisheyes is better.
>
> Peter M
>


Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

Bernhard Vogl-2
In reply to this post by Roger D Williams
Thanks for the detailed description which was very interesting to read - and the result is really excellent - would like to see it on my Zalman  ;-)

I can understand that the complex workflow is necessary for repeatable results. However, i would be interested to hear if you think that an excellent optimization of L/R indepentently could be a viable solution. This is what i try to do with the high resolution stereos - and i blame the wrong depth flaws my impatience to do a real good optimization. Would like to hear your thoughts about it!

Best regards
Bernhard

-------- Original-Nachricht --------
> Datum: Tue, 21 Dec 2010 19:30:54 +0900
> Von: "Roger D. Williams" <[hidden email]>
> An: [hidden email]
> Betreff: Re: [PanoToolsNG] Re: Hyperstereo pole panorama  -- anaglyph

> On Tue, 21 Dec 2010 19:09:03 +0900, panovrx <[hidden email]> wrote:
>
> >
> >
> >
> > --- In [hidden email], Andras Frenyo <andras@...> wrote:
> >>
> >> Super.
> >> The height, the spot, the separation - all has worked pretty perfect  
> >> here.
> >>
> >> How many images did you use to stitch the panos.  Can I assume not all
> >> and the same for both?
> >>
> >> Thanks.
> >>
> >> On Dec 19, 2010, at 8:38 AM, [hidden email] wrote:
> >>
> >> > http://www.mediavr.com/artgallerystereoana.htm
> >>
> >>
> > It was a rather convoluted workflow but this method seemed to work well
> > with this subject.
>
> I found this both fascinating and extremely depressing. The result is
> great, but that it took so much expertise and detailed work means that
> I won't be able to produce anything nearly as good in this lifetime.
> I can't imagine that Sony will ever automate anything like THIS!
>
> Roger W.
>
> --
> Business: www.adex-japan.com
> Pleasure: www.usefilm.com/member/roger
>
>
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

enridp
In reply to this post by jrgen_schrader
Congratulations Peter! your 3D pano is awesome.
I have a question, why you are using 2 cameras? I mean, what are the advantages of using two cameras instead of one (Wim's method).
Because I thought that the advantage was less pictures, but 50 shots is the same number of pictures needed with one camera.

Regards!
Enrique

--- In [hidden email], "jrgen_schrader" <panorama@...> wrote:

>
> Thank you very much, Peter, for the detailed insight.
> Much appreciated.
>
> After I had made some more tests with Wim's single camera solution I wondered how it would be possible to get the panoramas from to cameras with an offset leveled and parallelized so that they would make a useable stereo pair. Your description helped a lot.
>
> Jürgen
>
>
>
> --- In [hidden email], "panovrx" <panovrx@> wrote:
> >
> >
> >
> >
> > --- In [hidden email], Andras Frenyo <andras@> wrote:
> > >
> > > Super.
> > > The height, the spot, the separation - all has worked pretty perfect here.
> > >
> > > How many images did you use to stitch the panos.  Can I assume not all and the same for both?
> > >
> > > Thanks.
> > >
> > > On Dec 19, 2010, at 8:38 AM, [hidden email] wrote:
> > >
> > > > http://www.mediavr.com/artgallerystereoana.htm
> > >
> > >
> > It was a rather convoluted workflow but this method seemed to work well with this subject.
> > First I trimmed the L and R sequences, head and tail, to enough matched pairs to cover 360 and a little bit more.
> > About 56 images for each side. Then I check the sync by looking at action in the pairs. Then I loaded the L shots
> > and masked them in PTGui 9 so that only an area above the horizon and about 60 degrees wide was being used
> > for point finding. Then I calibrated the L shots to see if I could get a good solution. When it looked ok I levelled it.
> > Then I loaded the first of the R hand shots as well. I calibrated its yaw, pitch and roll against the first of the L shots.
> > Then I loaded the rest of the R shots.
> >
> > Now I know what roll, pitch and yaw of the first L shot will be level. What fov and a,b, c of the fisheye lens will give
> > a good stitch, and what roll, pitch and yaw of the first R shot will make it parallized to the first L shot.
> >
> > So now I set all the L shots to the same roll,pitch and yaw as the first L  shot and all the R shots to the same as
> > the first  R shot. So now all the image pairs are parallelized (like stereo pairs with the same roll and tilt and aligned
> > at the same infinity points),
> >
> > Then I convert all the fisheye images to 20 degree wide, 180 degree high equi strips using these settings.
> >
> > Then I load the equi L strips and optimize them (after masking the foreground for point finding)
> >  and stitch the final L panorama with Smartblend blending.
> > I output individual 360 panoramas for each L strip also.
> >
> > Then I make a template of this solution and apply it to the R strips and output the finished R panorama and individual
> > panoramas for each R strip.
> >
> > Smartblend will have blended differently in different action parts of the scene so you now have to fix those
> > (in a couple of places) with pasted bits from the individual  strip panoramas you have outputted.
> >
> > Then I load the finished L and R stitched and fixed panoramas into PTGui and level them again each other and finetune
> > the scene levelling.
> >
> >
> >
> > You will think this has too many steps and I could have stitched directly from the fisheyes  but I think the solution I got
> > with the second generation images (equi strips from the parallelised calibrated fisheyes) was quite a bit more accurate
> > than what I got directly from stitching the first L fisheye test stitch. But sometimes directly from the fisheyes is better.
> >
> > Peter M
> >
>


Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

panovrx-2

--- In [hidden email], "enridp" <enridp@...> wrote:
>
> Congratulations Peter! your 3D pano is awesome.
> I have a question, why you are using 2 cameras? I mean, what are the advantages of using two cameras instead of one (Wim's method).
> Because I thought that the advantage was less pictures, but 50 shots is the same number of pictures needed with one camera.
>
> Regards!
> Enrique
>
Single camera methods are only good for static scenes

PeterM

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

Ignacio Ferrando Margelí-3
In reply to this post by panovrx-2
Peter,

 that's really amazing!!

 I NEED to try that!!!!


regards



El 19/12/2010, a las 00:35, panovrx escribió:

> http://www.mediavr.com/artgallerystereoana.htm
>
> from about 4.5m with 50 shots in 18 seconds on a rotating twin 5D/10.5mm rig with camera separation 25cm
> ...like
> http://www.youtube.com/watch?v=EwzCMPeS6JE
> except with the cameras further apart and hanging down from a bar and rotating slower
>
> When you are far enough away from the nearest feature, like high on a pole, quite large disparity pano pairs can be stitched
>
> PeterM
>
>

----------------------------------------------------------

Ignacio Ferrando Margelí

Abaco Digital

http://www.abaco-digital.com







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



------------------------------------

--
<*> Wiki: http://wiki.panotools.org
<*> User Guidelines: http://wiki.panotools.org/User_Guidelines
<*> Nabble (Web) http://n4.nabble.com/PanoToolsNG-f586017.html
<*> NG Member Map http://www.panomaps.com/ng
<*> Moderators/List Admins: [hidden email]
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/PanoToolsNG/

<*> Your email settings:
    Individual Email | Traditional

<*> To change settings online go to:
    http://groups.yahoo.com/group/PanoToolsNG/join
    (Yahoo! ID required)

<*> To change settings via email:
    [hidden email]
    [hidden email]

<*> To unsubscribe from this group, send an email to:
    [hidden email]

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

enridp
In reply to this post by panovrx-2
Hi Peter, you are right, but 50 pictures and 18 seconds it's the time needed for one camera, that's why I'm so confused...
What's the difference of taking 50 pictures in 18 with two cameras and 50 pictures in 18 seconds with one camera?

Regards!!
Enrique.

--- In [hidden email], "panovrx" <panovrx@...> wrote:

>
>
> --- In [hidden email], "enridp" <enridp@> wrote:
> >
> > Congratulations Peter! your 3D pano is awesome.
> > I have a question, why you are using 2 cameras? I mean, what are the advantages of using two cameras instead of one (Wim's method).
> > Because I thought that the advantage was less pictures, but 50 shots is the same number of pictures needed with one camera.
> >
> > Regards!
> > Enrique
> >
> Single camera methods are only good for static scenes
>
> PeterM
>


Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

panovrx-2

--- In [hidden email], "enridp" <enridp@...> wrote:
>
> Hi Peter, you are right, but 50 pictures and 18 seconds it's the time needed for one camera, that's why I'm so confused...
> What's the difference of taking 50 pictures in 18 with two cameras and 50 pictures in 18 seconds with one camera?
>
> Regards!!
> Enrique.
>
 
Sorry I dont understand your question then. 50 shots each per camera, 100 altogether, with my panorama, if that makes it clearer.

PeterM

Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

onezebra1
This information is very much appreciated as I'm going to give this 3D shooting a try.
My second Canon 5D-2 should be arriving tomorrow and I already have 2 Nikon 10.5mm lens, but I'm thinking of trying it with 2 Canon 15mm mounted diagonally.

Roger Berry


Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

Wim Koornneef
In reply to this post by enridp
Hello Enrique,

Peter is right, a single cam solution can only be used for a static or frozen scene.

The main difference in using a single cam or a dual cam solution is not about the shooting time, or the number of images, or the rotation speed, or the fact that 2 cameras are used, no it is about the time shift between the stitched and aligned left/right images.

Depending on the shooting params (rotation speed, hFOV, out of NPP forward lens shift) with a single cam solution it can happen that the time shift can be many seconds between the left and the right image.
For static objects a few seconds time shift is never an issue but as soon as there are moving objects in the scene you get super ghosting when the left and right images are used for 3D.
Each of the individual left and right images can be looking great but when you use them together for 3D you get "super ghosting".

For this reason you always have to use a dual cam solution in a dynamic scene and even then it can be very hard to get rid of "normal ghosting" between the images of the same camera , to tackle this problem Peter is using Smartblend with relatively small vertical strips. Strips are used and not the full equirectangulars to make sure that Smartblend will only work on 2 neighbouring images and not on a large series of images to prevent time shift errors and even then it can be needed to repair errors because Smartblend can use different blending seems for the left and right images.

I am sure Peter had a bit of luck with this pano because a rotation speed of 18 seconds is relatively slow, in general "faster" is always "better".

Wim
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: Hyperstereo pole panorama -- anaglyph

no way


--- In [hidden email], Wim Koornneef <wim.koornneef@...> wrote:
>
>
> Hello Enrique,
>
> Peter is right, a single cam solution can only be used for a static or
> frozen scene.

Hello,
 Not entirely true. There is always the Loreo lens in a cap solution.
K

12
Loading...