[ITK Community] Fwd: Problems with slice extraction from a 3D image for directions different from z

[Simon Wimmer]

Hi all, hi Luis,

while I tried to put together a minimal example, I noticed that I had
forgot to call Update() before printing the size of the output image.
Now I understood that I have to do a manual resizing of the output image to
get a more reasonable visual result representing the real image spacing.
(Of course image spacing is just meta data, but I hadn't realized that
before)
Everything works fine for me now!
Thanks for your help!

Simon


2014-03-09 21:32 GMT+01:00 Luis Ibanez <luis.ibanez at kitware.com>:

> Simon,
>
> I'm looking at the code of the filter and there are
> not many ways for it to generate unit spacing.
>
> Output image spacing is computed in line 195:
>
> https://github.com/InsightSoftwareConsortium/ITK/blob/master/Modules/Core/Common/include/itkExtractImageFilter.hxx#L195
>
> Something else is going on with your code...
>
> Could you reduce it to a minimal .cxx that displays
> this behavior, and share it in the list ?
>
> In that way we will able to attempt to replicate the problem
>
>
>     Thanks
>
>
>        Luis
>
>
>
> On Sat, Mar 8, 2014 at 8:42 PM, Simon Wimmer <wimmersimon at gmail.com>wrote:
>
>> Hi all,
>>
>> as a follow up, I just checked the spacings in the original image and the
>> extracted slice.
>>
>> reader->GetOutput()->GetSpacing()
>> gives
>> [0.747159, 0.747159, 6].
>>
>> But
>> filter->GetOutput()->GetSpacing()
>> gives
>> [1, 1],
>> which explains the "loss of spacing information" I observed.
>> I guess if the spacing was preserved the filter->
>> SetDirectionCollapseToIndentity() setting would give me the desired
>> slice images.
>> Any tips on how I could track down the problem here?
>>
>>
>> Simon
>>
>>
>> 2014-03-08 22:10 GMT+01:00 Luis Ibanez <luis.ibanez at kitware.com>:
>>
>> Hi Simon,
>>>
>>> On Sat, Mar 8, 2014 at 3:56 PM, Simon Wimmer <wimmersimon at gmail.com>wrote:
>>>
>>>>
>>>> My direction matrix is simply the identity.
>>>> So I would guess that in this case filter->
>>>> SetDirectionCollapseToIndentity() and filter->
>>>> SetDirectionCollapseToSubmatrix() should do the same?
>>>>
>>>>
>>> It should...yes... I think.
>>>
>>> mm, maybe we need to take a closer look at the way the Submatrix is
>>> being composed:
>>>
>>>
>>> https://github.com/InsightSoftwareConsortium/ITK/blob/master/Modules/Core/Common/include/itkExtractImageFilter.hxx#L223
>>>
>>>
>>> https://github.com/InsightSoftwareConsortium/ITK/blob/master/Modules/Core/Common/include/itkExtractImageFilter.hxx#L188
>>>
>>>
>>>
>>>
>>>>  Taking out  InPlaceOn() sounds reasonable.
>>>>
>>>
>>> OK
>>>
>>>
>>>>  Regarding the spacing issue my intuition would have been that the
>>>> extraction of course destroys the information for the projection direction,
>>>>  but should preserve it for the other ones. Is that correct?
>>>>
>>>
>>> Yes, that's correct.
>>>
>>> So, if the initial 3D image had spacing values
>>>
>>>         1.7,  2.9,   3.1
>>>
>>> in X, Y, Z respectively,
>>>
>>>
>>> and you collapse Z,
>>> you should end up with a 2D image with spacing values
>>>
>>>          1.7,   2.9
>>>
>>> in X and Y respectively.
>>>
>>>
>>>   HTH
>>>
>>>
>>>       Luis
>>>
>>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/community/attachments/20140310/3601d40e/attachment-0002.html>