[Insight-users] Multi Modality Multi Resolution Doubt : Panoramas,
Mosaics, Stitching
Luis Ibanez
luis.ibanez at kitware.com
Sun, 29 Feb 2004 22:33:37 -0500
Hi Arash,
I'm not aware of public programs for building mosaics via registration.
However you may want to join forces with other ITK users who have
similar interests. Anybody working on Microscopy and Teledetection
will benefit from a framework for registering slices as mosaics.
The basic elements for performing this task are already in the toolkit,
what we should consider adding is a 'manager' class that will take a
set of images, with a set of pair specifications (e.g. image 15 has a
common border with image 21...) and then performs iterative refinements
of registration by computing transforms between pairs. It will be a class
with a role similar to the one played by the Multi-Resolution registration
class.
If you are interested in pursuing a collaboration with other users, we
will be happy to put an skeleton for this class and let it evolve accoding
to your needs.
Please let us know,
Thanks
Luis
--------------------------
Arash Jahangir wrote:
>Thanks Luis. Do you happen to know of any public programs/examples that I
>could look at to get more familiar with the process?
>
>TIA,
>Arash
>
>PS Forget Billy Crystal; they should have gotten you to host the show :-)
>
>
>----- Original Message -----
>From: "Luis Ibanez" <luis.ibanez at kitware.com>
>To: "Arash Jahangir" <arash at vije.ca>
>Cc: "ITK" <insight-users at itk.org>
>Sent: Sunday, February 29, 2004 5:04 PM
>Subject: Re: [Insight-users] Multi Modality Multi Resolution Doubt
>
>
>
>
>>Hi Arash,
>>
>>The 20 to 30 degrees that a registration method can correct for
>>are assumed to be *after* you provide an initialization.
>>
>>If you have images with a relative rotations that is larger than
>>30 degrees, but you can provide an initialization by for example
>>using landmarks, then the actual rotation that the registration
>>method has to correct is only the residual rotation error left after
>>the initialization.
>>
>>The problem of reconstructing a panorama is very similar
>>to the problem of constructing mosaics in teledetection or
>>stitching images in microscopy. They can be solved with
>>the current registration framework in ITK.
>>
>>You will find usefult to take advantage of the masking
>>capabilities that have been added to some metrics.
>>
>>The extra dificulty in mosaic is that you don't only need
>>to register two images but a set of N images in subsets.
>>Therefore you may have to iterate among the pairs of
>>neighbor images and redistribute the geometric transform
>>over the entire set of images.
>>
>>
>>Regards,
>>
>>
>> Luis
>>
>>
>>
>>
>>----------------------------
>>Arash Jahangir wrote:
>>
>>
>>
>>>Hi Luis,
>>>Would this range of 20-30 degrees be still the limiting factor if I had
>>>
>>>
>some
>
>
>>>user input that could roughly identify matching regions between the
>>>
>>>
>images
>
>
>>>as a starting point? To put this question in perspective, let me explain
>>>what I am after.
>>>
>>>I have a set of photos that form a "panorama" with some overlap. Because
>>>the subject was relatively close to the camera, the rotational distortion
>>>
>>>
>in
>
>
>>>the overlapping region is quite obvious (as compared to say a panorama of
>>>the Rocky Mountains). I am trying to write a program that would merge
>>>
>>>
>all
>
>
>>>these photos with each other. Do you have any suggestions for how I
>>>
>>>
>should
>
>
>>>go about this?
>>>
>>>thanks,
>>>Arash
>>>
>>>----- Original Message -----
>>>From: "Luis Ibanez" <luis.ibanez at kitware.com>
>>>To: "CSPL" <hyd2_affable at sancharnet.in>
>>>Cc: <insight-users at itk.org>
>>>Sent: Saturday, February 28, 2004 10:55 AM
>>>Subject: Re: [Insight-users] Multi Modality Multi Resolution Doubt
>>>
>>>
>>>
>>>
>>>
>>>
>>>>Hi Abdul,
>>>>
>>>>A reasonable registration proces will correct for
>>>>rotations of about 20 to 30 degrees. Expecting
>>>>it to correct for image orientation (axial, coronal
>>>>saggital) is unrealistic and actually dangerous.
>>>>
>>>>Registration is an optimization problem that is solved
>>>>with a cost function with multiple extrema. There is
>>>>no magic here, if you don't initialize the registration
>>>>close to the optimal region, you can end up anywhere,...
>>>>and don't have to much right to complain.
>>>>
>>>>
>>>>There are filters in ITK for correcting Flipping and
>>>>orienting images:
>>>>
>>>>
>>>>
>>>http://www.itk.org/Insight/Doxygen/html/classitk_1_1OrientImageFilter.html
>>>
>>>
>>>>http://www.itk.org/Insight/Doxygen/html/classitk_1_1FlipImageFilter.html
>>>>
>>>>
>>>>
>>>>
>>>>
>>http://www.itk.org/Insight/Doxygen/html/classitk_1_1ReflectImageFilter.html
>>
>>
>>>
>>>
>>>>This is something that has to be done very carfully since
>>>>you could easily switch left-right on the patient image
>>>>and lead a surgeon to operate on the wrong side.
>>>>
>>>>
>>>>So, to answer your questions explicitly:
>>>>
>>>>1) Yes, you must set the orientation to be the same
>>>>2) No, ITK will not do that for you.
>>>>3) Yes, you need to supply parameters.
>>>>
>>>>
>>>>
>>>>Regards,
>>>>
>>>>
>>>> Luis
>>>>
>>>>
>>>>-----------------
>>>>CSPL wrote:
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>>Hi,
>>>>>
>>>>>I need to do some registration of SPECT[brain] with MRI[segmented
>>>>>brain]. So i have choosen Multi Modality Multi Resolution Registration
>>>>>and stuck with some problems. I am not getting the final results upto
>>>>>the mark. Later i found that my MRI is in Coronal View and the SPECT is
>>>>>in Transverse View. I want you to clarify in this point that
>>>>>
>>>>>1. Do we need to set the orientation of both the volumes same? or
>>>>>2. Does ITK will itself look for this and REGISTER? or
>>>>>3. Do we need to supply any parameters to get REGISTRATION even in this
>>>>>situation?
>>>>>
>>>>>Please reply.
>>>>>
>>>>>Thanks in Advance
>>>>>S. Abdul Rahim
>>>>>
>>>>>
>>>>>
>>>>>
>>>>_______________________________________________
>>>>Insight-users mailing list
>>>>Insight-users at itk.org
>>>>http://www.itk.org/mailman/listinfo/insight-users
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>_______________________________________________
>>>Insight-users mailing list
>>>Insight-users at itk.org
>>>http://www.itk.org/mailman/listinfo/insight-users
>>>
>>>
>>>
>>>
>>>
>>
>>
>>
>>
>>
>>
>
>_______________________________________________
>Insight-users mailing list
>Insight-users at itk.org
>http://www.itk.org/mailman/listinfo/insight-users
>
>
>