[Insight-users] Portal and DRR image registration

[Quang Tran]

Hello everybody,

I have problems when perform Portal and DRR images registration (now I only
test with 2D-Case). Portal image is very noisy, low contrast and have a
wide range of intensity (from 0 to 60000), so that when I tried to run some
algorithms of ITK, the results very very bad!

I know that Portal and DRR are in different modality, so that if I use a
intensity-based method of registration I should use Mutual Information
metric, however the results not good.

If I use a feature-based method, I must extract the features of images and
matching them, the features can be here is: harris corner, SIFT, SUFT,
MSER... However with Portal image it is very difficult to detect these
keypoints.

Now I am trying using ICP to match contour extracted from two images (using
VTK), the results is better than other way but still isn't my expectation.

I have read the file
http://www.medphys.mcgill.ca/presentations/keyvan2004.pdf and I 'm going to
use FFT-method to do this task.

Could anyone give me some advises with my problem?

Thanks in advance.

/Quang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20120416/7089910f/attachment.htm>