[Insight-users] spatial-object to pointset registration (still big problems)

[Boettcher, Dr. Peter]

Thank you to all how respond to my last email

 

However I still got stuck developing my spatial-object to point-cloud registration.

I moved to another approach using a spatial object (cylinder) and a point cloud generated by picking points in 3D within vtk.

 

I convert the point cloud using itkPointSetToImageFilter, getting a binary image of the point cloud, so I can plug it in the model-to-image registration example from the itk book. However, when I look at the resulting image volume within vtk, I get very strange pictures. I think that is due to the parameters: spacing, origin and size, used for the PointSetToImageFilter. This it what I thought it should be:

 

The volume within vtk I pick the points on is 50mm x 50mm x 50mm. Originally the volume was represented by voxels of 0.25mm x 0.25mm x 0.5mm. 

So this is why I choose the following size fore the itkPointSetToImageFilter

50mm/0.25 -> 200 pixel (x)

50mm/0.25 -> 200 pixel (y)

50mm/0.25 -> 100 slices (z)

As for the spacing I took the spacing from the original image volume

0.25 - 0.25 - 0.5.

The origin is defined as (0,0,0), indicating the lower left corner of the volume. 

 

Defining itkPointSetToImageFilter for the registration in the way I did should result in a transformation I could transform directely within my vtk-volume-space (hope so) !?

 

Unfortunately the filtering of the point cloud to an image volume does not work as I expected. 

Maybe someone used a similar approach and can show me how to do it the right way?

 

Regards, Peter.

 

PS: does the cylinder spatial object respresents a filled cylinder or does it represent an empty one?

 
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/insight-users/attachments/20051003/71866f65/attachment.html