[Insight-users] Any ideas and suggestions about section finding using ITK?

[Yuanxin Zhu]

Hello Everyone,

I'm working on a section finding problem as illustrated by the two
images in the attachment (The JPEG image is the original data.  The PNG
image is one of the current results.)  My goal is to develop a
program to automatically detect/recognize zero/one/multiple sections in
each image and then register them.

So far, I have used itk::WatershedImageFilter,
itk::RelabelComponentImageFilter, and BinarythresholdImageFilter to
generate binary masks for each sections found using a priori
information about the object size. An example original image and one of
the masks generated for an image were attached with this e-mail.

I have the following questions:

1) After obtaining the binary masks, how could I cut the the
sections/regions  out of the original image? Is there a existing ITK
class capable of doing this?

2) There are images in my application where comparable objects show up but
are not the sections I want to extract. To filter out this kind of objects
or any other noise false alarms, I plan to do a second stage of process by
image classification. My question is that what kinds of features should be
calculated, e.g., shape moments? Is there existing ITK class capable of
calculating image features and then classifying images using those
features?

Any ideas or suggestions are highly appreciated.  Those can be implemented
using existing ITK classes are particularly welcome.

Thank you in advance.

Yuanxin

-------------- next part --------------
A non-text attachment was scrubbed...
Name: 55.obj3.png
Type: application/octet-stream
Size: 1020 bytes
Desc: 
Url : http://public.kitware.com/pipermail/insight-users/attachments/20040809/244a3782/55.obj3-0001.obj
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 55.jpg
Type: image/jpeg
Size: 26201 bytes
Desc: 
Url : http://public.kitware.com/pipermail/insight-users/attachments/20040809/244a3782/55-0001.jpg