[Insight-developers] One possible framework for ImageI/O...

[Stephen R. Aylward]

Hi Will-

You raise some good points, but for the record

1) In our WA, we said we'd do DICOM and MetaImage IO.   I asked Parag to
look into it, and this is what he came up with.   It is nice and
extensible and doesn't break other people's code.   Streaming isn't
initially addressed, but I think Parag has ideas on it.

2) Parag did send a notice to the Insight list several weeks ago, and
since he didn't get any complaints, he submitted.

3) His paper is quite specific about the class designs.   He wrote it to
document the classes.   I think this brings up the issue of how to
document a set of related classes that had been brought up by someone
else on the list a few days ago.

BUT - as I said - you raise good points.   Parag will move the work to
the interfaces section.

I understand that there has also been some concern about the
registration class hierarchy that we submitted. This too was my doing. I
asked Graham Gash, Luis Ibanez, and Sarang Joshi to come up with an
extensible hierarchy.   We left FEMs alone and tried to take Lydia's
work into consideration, but perhaps we should be thinking smaller for
this set of classes as well.

****** I guess my delimma has been trying to get stuff done for the
immediate work assignment while keeping extensibility and code re-use in
mind.   I didn't want the people I work with to check-in functions for
two file types, two registration methods, and a few other methods
without including support for other file types, other registration
methods, and other functions.  Piecemeal/function-level development
doesn't seem ideal, e.g., we'll end up with different APIs for different
methods, etc.

In hindsight, we should have been communicating more - perhaps this
applies to the algorithm developers in general.   I think the group
meetings will help...and you'll be able to get retribution on the slopes
if I decide to try snowboarding again...

Thanks for the guidelines - we will follow them with the statistics and
active surface classes we are about to (i.e., may) submit.

Stephen

-- 
===============================================
Stephen R. Aylward
Assistant Professor of Radiology
Adjunct Assistant Professor of Computer Science
http://www.cs.unc.edu/~aylward
aylward@unc.edu
(919) 966-9695