[Insight-users] Thanks, Another Question

[Luis Ibanez]

Hi Raheleh,


One of the powerful features of Open Source
software is that you can study the software
in order to fully understand what it does.


             "Glimpsing at the Source
              leaves no doubt"



Please look at the files:

Insight/Code/Algorithms/
   itkMultiResolutionPyramidImageFilter.h
   itkMultiResolutionPyramidImageFilter.txx
   itkRecursiveMultiResolutionPyramidImageFilter.h
   itkRecursiveMultiResolutionPyramidImageFilter.txx


You will find that these classes use the

             DiscreteGaussianImageFilter

to smooth the images, an then uses the


           ShrinkImageFilter

for doing subsampling.



      Regards,


             Luis


-------------------------------------------------------------------------
On Mon, Jun 21, 2010 at 1:34 AM, raheleh m <meshgolok at yahoo.com> wrote:

> Dear Mr. Ibanez
>
> Thank you so much for your attention. I am really grateful for your help.
> I am using 3 Resolution levels and the subsampling schedule is
> [8 8 1], [4 4 1], [2 2 1] for Moving Image and [4 4 1],[2 2 1], [1 1 1] for
> Fixed Image.
> It would be really helpful if you could please help me to understand what
> prepare pyramids do? I mean for down sampling, what kind of filters does it
> use? Can I improve the procedure by removing some of these filters or it
> would affect the results?
>
> Best Regards,
> Sincerely Yours
>
> Rahil Meshgolok
>
> ------------------------------
> *From:* Luis Ibanez <luis.ibanez at kitware.com>
> *To:* raheleh m <meshgolok at yahoo.com>
> *Cc:* itk <Insight-users at itk.org>
> *Sent:* Sun, June 20, 2010 5:30:20 PM
> *Subject:* Re: Thanks, Another Question
>
> Hi Rahil
>
> 1) Before you start measuring computation time,
>     please make sure that you configured ITK to
>     build for "Release". This is done by running
>     CMake and setting the variable
>
>                    CMAKE_BUILD_TYPE
>
> 2)  Regarding the time for building the pyramid,
>      170 seconds looks a bit too long...
>
>      How many resolution levels are you using ?
>      What is the subsampling schedule ?
>
>
>
>   Please let us know,
>
>
>        Thanks
>
>
>              Luis
>
> -------------------------------------------
> On Sun, Jun 20, 2010 at 8:28 AM, raheleh m <meshgolok at yahoo.com> wrote:
>
>> Dear Mr. Ibanez
>>
>> Thanks a lot for your attention and help. Actually, I have another
>> question and I would be really grateful
>> if you could please help me with this one too. I am using Multi Resolution
>> Registration method of itk, and my data size is 512*512*200. The prepare
>> pyramids function takes about 170 seconds, but the optimization loop is
>> fast, about 12 seconds. The total registration time is about 202 seconds. Is
>> this normal, or I miss something? How can I reduce the time of prepare
>> pyramids? I know it is a lot to ask and I appreciate it a lot.
>>
>> Thanks in advance,
>> Sincerely Yours
>>
>> Rahil Meshgolok
>>
>> ------------------------------
>> *From:* Luis Ibanez <luis.ibanez at kitware.com>
>> *To:* raheleh m <meshgolok at yahoo.com>
>> *Cc:* itk <Insight-users at itk.org>
>> *Sent:* Thu, June 17, 2010 8:40:41 PM
>> *Subject:* Re: Permute Order in MultiResMIRegistration
>>
>>
>> Hi Rahil,
>>
>>
>> Please read the documentation of the class at:
>>
>> http://public.kitware.com/Insight/Doxygen/html/classitk_1_1FlipImageFilter.html
>>
>>       "The image is flipped across axes for which array[i] is true."
>>
>>
>> http://public.kitware.com/Insight/Doxygen/html/classitk_1_1PermuteAxesImageFilter.html
>>
>>      "The elements of the array must be a rearrangement
>>       of the numbers from 0 to ImageDimension - 1.
>>
>>      "The i-th axis of the output image corresponds with
>>        the order[i]-th axis of the input image."
>>
>> If you pass [0 1 2] as parameter, then the output image will be identical
>> to the input image.  If you pass [1 0 2] then the output image will use as
>> X axis, what used to be the Y axis of the input image, and will use as
>> Y axis, what used to be the X axis of the input image.
>>
>>
>> Note that in general, you should NOT use the flip / permute
>> filters. Particularly not if you are using medical data.
>>
>>
>>      Regards,
>>
>>
>>           Luis
>>
>>
>> -------------------------------------------------------------------------
>> On Mon, Jun 14, 2010 at 8:12 AM, raheleh m <meshgolok at yahoo.com> wrote:
>>
>>> *Greetings!
>>> *
>>> I am using ITK Registration. I would be really grateful if you could
>>> please let me know the meaning of permute order and flip axes arrays. I mean
>>> when permute order is [0 1 2] what is the order of axes? what about [2 0 1]?
>>> Does the first array mean (x,y,z) and the second (z,x,y)? I have the same
>>> question about flip axes.
>>>
>>> Thanks in advance
>>> Rahil Meshgolok
>>>
>>>
>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20100621/dd50d882/attachment.htm>