How to patch ITK4 to CITK

[Mengqiu Tian]

Hi all,

Is here someone who know how to patch ITK4 to CITK?

Thx a lot for help :)

Mengqiu

[Phillip Ward]

This hasn't been explored at this point in time. I apologize for the delayed response, I have been away on leave for a few weeks.  There is one member of the team looking into this but it is not high on the priority list due to the interests of main user groups.

[Mengqiu Tian]

Hi Phillip,

Anyway, thx for replying.

I am looking forward of ITK4 in CITK, otherwise...I have to rewrite my program. :(

Mengqiu

[Phillip Ward]

That is a shame. May I ask a bit about how you're using CITK? What type of use are you getting out of it? Have you encountered many limitations or bugs? What is your application?

[Mengqiu Tian]

I am totally new about CITK. My supervisor want to do translate matlab programs into C++ in order to accelerate the image processing. And in order to achieve higher acceleration, he decided to use CUDA instead of CPU.  Actually  it is my fault, I wrote only a simple program which compose scalar images into a vector image, and I forget the version things, the head file which I use only exits in ITK4, not in ITK3.20. In most cases, we deal with raw images by writing a mhd file.

[Phillip Ward]

What type of operations is he doing? I'd considering scaling back to ITK3.20, CITK will not be being adapted to ITK4 soon.

[Mengqiu Tian]

I would take your advise and I also looks for another head file to implement it. Actually, I am not quite clear about the whole project yet. I just started to familiar with CITK by writing some sample examples. He focus on image coding, but I this our project will do something about medicine image segmentation and registration. He is working on ITK3.20, it is only my fault that choosed a wrong version. But I think the architecture of ITK4 is clearer than ITK3.20. Hope it helps.

[Phillip Ward]

ITK4 is a substantial shift, and may take a little while for bugs to be ironed out. I've spoken to another guy in the group who has a bit more time on his hands and he is trying to get CITK compatible with it, no promises though.

[Mengqiu Tian]

That would be fine, although not promises. ;P

Thanks a lot! And I will look forward the new one.

Mengqiu

[Taylor Braun-Jones]

Hi Phillip,

I'm curious if there has been any progress on the ITKv4 port in the last year. Is there a work-in-progress branch I could check out and potentially help out with? We use ITK within a larger toolkit (MITK) and must upgrade to the 4.x series when 3.20 support is deprecated in MITK.

Thanks!

Taylor

[Phillip Ward]

Hi Taylor,

The work was abandoned and I do not believe the original work was ever associated with the SVN.

I'll be moving back into image processing and CUDA in the next 6 months which could see some time put back into CITK. My understanding was that ITK was putting GPU support in themselves (using the CITK design) into their new release, so it seemed unnecessary to proceed with CITK. Is this not the case?

If you want to have a crack at CITK for ITKv4 you are more than welcome. I'll help in any way I can.

Cheers,

Phil

[Taylor Braun-Jones]

Yes, that is the case that the new 4.x series added GPU support but it only
provides support for OpenCL. Several of our GPU kernels are implemented in
CUDA because it runs faster, more featureful, and has a much larger
community of libraries and tools than OpenCL.

My sense is that the patch to ITKv4 should not be insurmountable. For now
we are fine with ITK 3.20, I mostly wanted to get a feeling for how much
effort we'd be facing when it comes time to make the transition to 4.x.

- Taylor

P.S. - Thanks for a great addition to ITK!

[Phillip Ward]

Hey Taylor,

It is probably better to create a new CITK on top of ITKv4, rather than trying to modify the existing CITK to fit it. Since it has support, a flag in the build environment to specify OpenCL or CUDA, along with predicates to invoke the right compiler and preprocessor defines for alternate code paths should be enough. Perhaps a few weeks work full time if one knew what they were doing.

Cheers,

Phil