users > "x" vs. non-"x" tools and fast registration
Showing 1-1 of 1 posts
Display:
Results per page:
Feb 10, 2011  11:02 PM | Torsten Rohlfing - Google LLC
"x" vs. non-"x" tools and fast registration
Hi Dominique:

Q1 - for "reformatx" you have no choice. The plain "reformat" tool has been retired a few releases back. For the other two, the plain versions are currently more predictable and interface-stable, so I would suggest using "registration" and "warp".

Q2 - to make things fast, it is best to simply disable the optimization. For "registration" you should be able to simply set the number of DOFs to zero by adding "--dofs 0" to the command line. For "warp" I would suggest using the "--f-threshold" parameter. Specifically, using "--f-threshold 1" should virtually disable optimization. If it does not, try larger values, and just in case my memory is totally off here, try zero also.

Sorry for delay and vagueness - I am on vacation and thus slightly handicapped.

Best,
Torsten

PS: as usual I will post this reply to the CMTK users forum on NITRC also; feel free to use the forum for future inquiries ;)

On 2/8/2011 8:08 AM, Dominique Belhachemi wrote:
> Hi Torsten,
>
> In our EMSegmenter pipeline we are using the following three CMTK commands:
> registration
> warp
> reformat
>
> I think I asked this already, but I cannot find the answer anymore.
> Is it better to use the 'x' version of those commands, like registrationx, warpx, reformatx?
>
> For test purposes I do need some parameters to force those tools to execute pretty fast. The registration accuracy is completely unimportant, all I wanna do is to test our pipeline with preregistered images.
>
> Do you have those parameter available?
>
> Thanks
> Dominique
>