[Insight-users] Re: deformable registrations

[Luis Ibanez]

Hi Silvester,

Your deformed image looks great.

For ways of fine tunning the wrapping you may want to look at
the BSplineTransform which will use a grid of landmarks over
the images.

Another way to go is to use the KernelTransforms directly
but using an optimizer that *do not* require derivatives
(sing the Jacobians are not yet implemented in the transforms).
In this context the paramters would be the coordinates of the
target landmarks, because by moving the target landmarks you
control the entire deformation field.

An iteresting optimizer to try is the OnePlusOneEvolutionary
optimizer.

http://www.itk.org/Insight/Doxygen/html/classitk_1_1OnePlusOneEvolutionaryOptimizer.html



Regards,


    Luis



--------------------------
Silvester Czanner wrote:

> Luis,
> 
> thank you again for your help with the deformable registration. The 
> example you created is very good and it helped me a lot. I also used 
> this example to register 2 mouse embryos images and it worked fine. I am 
> sending you an example of deformed image.
> 
> There is another question, I do not know, how to answer.
> Is there any possibility to tune the warping process? My guess is, that 
> right now the warping process is setup with some moderate parameters, 
> but I can not figure out, what kind of parameters are there and how to 
> change it.
> 
> Thank you,
> 
> Silvester
> 
> Luis Ibanez wrote:
> 
>>
>> Silvester,
>>
>> Please find attached the results of
>> a first attempt to match your images.
>>
>> The file "landmarks.txt" contains
>> coordinates of corresponding point
>> from idt1 and idt2.
>>
>> By passing this file as argument
>> to the example
>>
>>    Insight/Registration/
>>               LandmarkWarping2.cxx
>>
>> I generated the image idt2deformed.png
>>
>> and using the .tcl scripts produced
>> the checkerboard pattern images "before"
>> and "after".
>>
>> you may have to fine tune the values
>> of this landmarks, and eventually
>> add more landmarks in other cell structures
>> that you want to match.
>>
>> Once you find a satisfactory initalization
>> you can combine the code of the LandmarkWarping
>> example with the code in DeformableRegistration2
>> in order to use the landmark deformation field
>> as an initialization of the DemonsRegistration
>> filter.
>>
>>
>> Please let us know if you find any problem.
>>
>>
>> Thanks
>>
>>
>>      Luis
>>
>>
>>
>> -------------------------------
>> Silvester Czanner wrote:
>>
>>> Thank you Luis,
>>>
>>> Silvester
>>>
>>> Luis Ibanez wrote:
>>>
>>>> Hi Silvester,
>>>>
>>>> I haven't had a chance to look at
>>>> registering your images yet.
>>>>
>>>> Hope to find some time to do it
>>>> this week.
>>>>
>>>> Regards,
>>>>
>>>>
>>>>    Luis
>>>>
>>>>
>>>> --------------------------
>>>> Silvester Czanner wrote:
>>>>
>>>>> Hi Luis,
>>>>>
>>>>> I was wondering, if you already tried to register the two images I 
>>>>> sent you.
>>>>>
>>>>> Thank you,
>>>>>
>>>>> Silvester
>>>>>
>>>>> Luis Ibanez wrote:
>>>>>
>>>>>> Hi Silvester,
>>>>>>
>>>>>> I'll need to play a bit with your images.
>>>>>>
>>>>>> The seems to have very large deformations
>>>>>> as well as a lot of not matching structures.
>>>>>> (structures that are in one image and not
>>>>>> in the other)
>>>>>>
>>>>>> I would try using the KernelBase Spline
>>>>>> as a deformable transform...
>>>>>>
>>>>>>
>>>>>> Regards,
>>>>>>
>>>>>>
>>>>>>    Luis
>>>>>>
>>>>>>
>>>>>> -------------------------
>>>>>>
>>>>>> Silvester Czanner wrote:
>>>>>>
>>>>>>> Luis,
>>>>>>>
>>>>>>> during last few days I tried to register 2 images (I am sending 
>>>>>>> you as attachments) using several ITK methods. Unfortunately, I 
>>>>>>> was not successful. I was wondering if there is some way in ITK 
>>>>>>> to register those two images. I was thinking about Procrustes 
>>>>>>> analysis, but I did not find such a method implemented in ITK. 
>>>>>>> Could you, please help me? Thank you very much.
>>>>>>>
>>>>>>> Silvester
>>>>>>>
>>>>>>>
>>>>>>> ------------------------------------------------------------------------ 
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> ------------------------------------------------------------------------ 
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>
> 
> 
> ------------------------------------------------------------------------
>