[Insight-users] Additional Info: Problems dewarping a fluoroscopic image

[Boettcher, Dr. Peter]

Dear Karthik

Thank you very much für your help. 


This is what the meta-image header of the deformation field looks like:

ObjectType = Image
NDims = 2
BinaryData = True
BinaryDataByteOrderMSB = False
CompressedData = False
TransformMatrix = 1 0 0 1
Offset = 0 0
CenterOfRotation = 0 0
ElementSpacing = 1 1
DimSize = 768 576
AnatomicalOrientation = ??
ElementNumberOfChannels = 2
ElementType = MET_FLOAT
ElementDataFile = kalip_grey_Deformtionfield_ReciprocalSpline_new.raw

Changing the scale mode to "Vector" results in arrows all oriented from left to right
http://www.bilder-upload.eu/show.php?file=KXzhwA4RAGTcTpD.jpg

However the data range of my deformation file is [-67.2714, 67.4893],[-61.8011, 60.6215].

Peter.


-----Ursprüngliche Nachricht-----
Von: Karthik Krishnan [mailto:karthik.krishnan at kitware.com] 
Gesendet: Sonntag, 13. April 2008 22:03
An: Boettcher, Dr. Peter
Cc: insight-users @ itk. org
Betreff: Re: [Insight-users] Problems dewarping a fluoroscopic image

On Sun, Apr 13, 2008 at 2:43 PM, Boettcher, Dr. Peter <Boettcher at kleintierklinik.uni-leipzig.de> wrote:
>
> I have serious problems dewarping a fluoroscopic image.

> Resampling of my fluo-image is done as it is shown in 
> LandmarksWarping2 example. I read in the landmarks, my fluo-image and 
> compute the dewarped fluo image. However what I get is not what I 
> expected. The image is less distorted than the original fluo-image, 
> but one can see that there is still marked distortion within the image.

I looked at the images and it does look odd. The TPS should've at least matched the landmarks perfectly in your case.

> I tried different kernels, I tried whether it would make any 
> difference using itkWarpImageFilter or itkResampleImageFilter,

Both should yield nearly the same results. WarpImageFilter using the deformation field or the ResampleImageFilter with the resulting TPS transform.
>
> I tried to look onto the deformation field - however the *.mda image 
> which is generated using the code of the deformable registration 
> examples does not make sense when I try to put some glyphs on it using 
> ParaView. I suppose that only the x-values of the vectors provided within the mda-file are used.

I looked at the paraview screenshot. Is this Paraview3 ?

It looks like you are orienting the vector glyphs simply by the scalar magnitude of the deformation.

Make sure you've checked the "Orient" flag (see properties panel of the Glyph filter) and make sure the scale mode is "Vector".

How are you loading the image into paraview. Is it a meta-image with 2 or 3 components ?


--
karthik