"Spherical autostitching"

[Joachim Wesner]

Hi List,

what is the status about (open) ImageJ plugins (or hints for algos) for
autostitching?

In my case, I would have the special problem that I want to stitch many
(even noisy!) images of a close full sphere surface (each image has only ~
1/4 of the full surfrace) at random orientations, i.e. even after
correction for the curvature it would not only be a translation but a
random orientation.

Anybody "been there, done that"? Hints?

Most sincerly

Mit freundlichen Grüßen / Best regards

Joachim


______________________________________________________________________
This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email 
______________________________________________________________________

[Johan Henriksson-2]

On Tue, Dec 22, 2009 at 2:18 PM, Joachim Wesner <
[hidden email]> wrote:

> Hi List,
>
> what is the status about (open) ImageJ plugins (or hints for algos) for
> autostitching?
>

trackem2 should always be mentioned here. I think it only does translation
and some warping (see site, haven't tried it yet).

then there is xuvtools but it was buggy when I tried it. they claim it to be
better than Imaris at least.


>
> In my case, I would have the special problem that I want to stitch many
> (even noisy!) images of a close full sphere surface (each image has only ~
> 1/4 of the full surfrace) at random orientations, i.e. even after
> correction for the curvature it would not only be a translation but a
> random orientation.
>
> Anybody "been there, done that"? Hints?
>

it's not obvious to me what you expect the output to look like. you want a
volume out of this or just aligned images? microsoft research has their
massive photo assembler which is the closest I recall to just aligning
images in 3d.

for volumes, well, there is the work on assembling SPIM images (rotated, a
little translated) for zebra fish but this assumes you have quite a number
of images. their algorithm requires that beads are placed a bit everywhere
to find the position.

/Johan

>
> Most sincerly
>
> Mit freundlichen Grüßen / Best regards
>
> Joachim
>
>
> ______________________________________________________________________
> This email has been scanned by the MessageLabs Email Security System.
> For more information please visit http://www.messagelabs.com/email
> ______________________________________________________________________
>

--
-----------------------------------------------------------
Johan Henriksson
PhD student, Karolinska Institutet
http://mahogny.areta.org  http://www.endrov.net

[Joachim Wesner]

Hi Johan,

thanks for your comments! I will look into it. Only to further clarify the
scope, it´s only a reconstruction of something similar to "surface
texture".

Image you have a collection of satellite images of the surface of the
earth, taken from pretty distant, so that the image is neither flat, but
the inclination of the surface relative to the point of view is
significant, but it´s usually also not a full semisphere in view. What the
problem would be, translated to this case, that you do not know where the
individiual satellite image was taken, you only know the distance, which is
usually fixed for a given sets of measurements, but in can be from any
orientation around the globe.

Actually, it would be a 3D reconstruction, but in my cases the "texture" is
small deviations from the exact sphere, so it acts like a "texture"

Actually it already exists for cases where the poistion of the "images" is
almost exactly known, you only do a fine tuning and some error compensation
in what would be the "satellite camera",
it´s calles "subaperture stitching interferometry", but IMHO nobody has
till now tried this for unkown orientations around a full sphere.

http://qedmrf.com/documentView.asp?docid=623
http://qedmrf.com/documentView.asp?docid=191


Sincerely, Merry Xmas and Happy New Year!

Joachim




                                                                           
             Johan Henriksson                                              
             <[hidden email]                                            
             G>                                                         An
             Gesendet von:              [hidden email]                
             ImageJ Interest                                         Kopie
             Group                                                        
             <[hidden email].                                       Thema
             GOV>                       Re: "Spherical autostitching"      
                                                                           
                                                                           
             22.12.2009 20:06                                              
                                                                           
                                                                           
              Bitte antworten                                              
                    an                                                    
              ImageJ Interest                                              
                   Group                                                  
             <[hidden email].                                            
                   GOV>                                                    
                                                                           
                                                                           




On Tue, Dec 22, 2009 at 2:18 PM, Joachim Wesner <
[hidden email]> wrote:

> Hi List,
>
> what is the status about (open) ImageJ plugins (or hints for algos) for
> autostitching?
>

trackem2 should always be mentioned here. I think it only does translation
and some warping (see site, haven't tried it yet).

then there is xuvtools but it was buggy when I tried it. they claim it to
be
better than Imaris at least.


>
> In my case, I would have the special problem that I want to stitch many
> (even noisy!) images of a close full sphere surface (each image has only
~
> 1/4 of the full surfrace) at random orientations, i.e. even after
> correction for the curvature it would not only be a translation but a
> random orientation.
>
> Anybody "been there, done that"? Hints?
>

it's not obvious to me what you expect the output to look like. you want a
volume out of this or just aligned images? microsoft research has their
massive photo assembler which is the closest I recall to just aligning
images in 3d.

for volumes, well, there is the work on assembling SPIM images (rotated, a
little translated) for zebra fish but this assumes you have quite a number
of images. their algorithm requires that beads are placed a bit everywhere
to find the position.

/Johan

>
> Most sincerly
>
> Mit freundlichen Grüßen / Best regards
>
> Joachim
>
>
> ______________________________________________________________________
> This email has been scanned by the MessageLabs Email Security System.
> For more information please visit http://www.messagelabs.com/email
> ______________________________________________________________________
>

--
-----------------------------------------------------------
Johan Henriksson
PhD student, Karolinska Institutet
http://mahogny.areta.org  http://www.endrov.net



______________________________________________________________________
This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email 
______________________________________________________________________

[Stephan Saalfeld]

Hi,

that's essentially panorama stitching---I do not know a solution for
ImageJ.  But first, I would give hugin a try:

http://hugin.sourceforge.net/

Hugin internally uses the PanoramaTools

http://panotools.sourceforge.net/

by which you should, hopefully, be able to estimate not only the
transformation but also the lens properties of your images and finally
stitch them with enblend

http://enblend.sourceforge.net/

It doesn't matter that Panorama Stitching related software thinks about
a rotating camera capturing the sphere from inside instead about a fixed
sphere captured by a moving camera.  The problem is basically the same
and just a question of an appropriate lens model.

Neither TrakEM2 nor xuvTools nor Stitching2D/3D nor TurboReg solve that
problem.  TrakEM2 and TurboReg work with up to affine transformations
plus lens-distortion, xuvTools and the Stitching plugins do translation
only.  Here, after compensating for the lens, the transformation we're
looking for is a subclass of a homography that is constrained by the
fact that the pictures come from a sphere.

Best,
Stephan



On Wed, 2009-12-23 at 13:27 +0100, Joachim Wesner wrote:

> Hi Johan,
>
> thanks for your comments! I will look into it. Only to further clarify the
> scope, it´s only a reconstruction of something similar to "surface
> texture".
>
> Image you have a collection of satellite images of the surface of the
> earth, taken from pretty distant, so that the image is neither flat, but
> the inclination of the surface relative to the point of view is
> significant, but it´s usually also not a full semisphere in view. What the
> problem would be, translated to this case, that you do not know where the
> individiual satellite image was taken, you only know the distance, which is
> usually fixed for a given sets of measurements, but in can be from any
> orientation around the globe.
>
> Actually, it would be a 3D reconstruction, but in my cases the "texture" is
> small deviations from the exact sphere, so it acts like a "texture"
>
> Actually it already exists for cases where the poistion of the "images" is
> almost exactly known, you only do a fine tuning and some error compensation
> in what would be the "satellite camera",
> it´s calles "subaperture stitching interferometry", but IMHO nobody has
> till now tried this for unkown orientations around a full sphere.
>
> http://qedmrf.com/documentView.asp?docid=623
> http://qedmrf.com/documentView.asp?docid=191
>
>
> Sincerely, Merry Xmas and Happy New Year!
>
> Joachim
>
>
>
>
>                                                                            
>              Johan Henriksson                                              
>              <[hidden email]                                            
>              G>                                                         An
>              Gesendet von:              [hidden email]                
>              ImageJ Interest                                         Kopie
>              Group                                                        
>              <[hidden email].                                       Thema
>              GOV>                       Re: "Spherical autostitching"      
>                                                                            
>                                                                            
>              22.12.2009 20:06                                              
>                                                                            
>                                                                            
>               Bitte antworten                                              
>                     an                                                    
>               ImageJ Interest                                              
>                    Group                                                  
>              <[hidden email].                                            
>                    GOV>                                                    
>                                                                            
>                                                                            
>
>
>
>
> On Tue, Dec 22, 2009 at 2:18 PM, Joachim Wesner <
> [hidden email]> wrote:
>
> > Hi List,
> >
> > what is the status about (open) ImageJ plugins (or hints for algos) for
> > autostitching?
> >
>
> trackem2 should always be mentioned here. I think it only does translation
> and some warping (see site, haven't tried it yet).
>
> then there is xuvtools but it was buggy when I tried it. they claim it to
> be
> better than Imaris at least.
>
>
> >
> > In my case, I would have the special problem that I want to stitch many
> > (even noisy!) images of a close full sphere surface (each image has only
> ~
> > 1/4 of the full surfrace) at random orientations, i.e. even after
> > correction for the curvature it would not only be a translation but a
> > random orientation.
> >
> > Anybody "been there, done that"? Hints?
> >
>
> it's not obvious to me what you expect the output to look like. you want a
> volume out of this or just aligned images? microsoft research has their
> massive photo assembler which is the closest I recall to just aligning
> images in 3d.
>
> for volumes, well, there is the work on assembling SPIM images (rotated, a
> little translated) for zebra fish but this assumes you have quite a number
> of images. their algorithm requires that beads are placed a bit everywhere
> to find the position.
>
> /Johan
>
>
>
> >
> > Most sincerly
> >
> > Mit freundlichen Grüßen / Best regards
> >
> > Joachim
> >
> >
> > ______________________________________________________________________
> > This email has been scanned by the MessageLabs Email Security System.
> > For more information please visit http://www.messagelabs.com/email
> > ______________________________________________________________________
> >
>
>
>
> --
> -----------------------------------------------------------
> Johan Henriksson
> PhD student, Karolinska Institutet
> http://mahogny.areta.org  http://www.endrov.net
>
>
>
> ______________________________________________________________________
> This email has been scanned by the MessageLabs Email Security System.
> For more information please visit http://www.messagelabs.com/email 
> ______________________________________________________________________