[Insight-developers] Re: [Insight-users] Image Coordinate Directions

[Leila Baghdadi]

Hello Mark

I agree that not everything is taking direction cosines into account but
I am pretty sure it will be if not now in the near future.

I am in the process of looking over some of my code and I will make sure
I will double check for direction cosines.

As for documentation, I know that Insight developers specifically Master
Luis will take your suggestions into consideration,


Leila

On Wed, 2006-08-03 at 09:34 -0500, Mark Foskey wrote:
> 
> Leila Baghdadi wrote:
> > Mark,
> > 
> > I have written a fair bit of imaging code and I can tell you ignoring
> > direction cosines has always got me into trouble! 
> 
> Well, that's what I thought too, but my point was that everybody else, 
> as far as I can tell, *is* ignoring the direction cosines.  At least all 
> the code in InsightApplications.  And, judging by an earlier thread, at 
> least one of the filters doesn't even propagate that information, 
> although that may have been fixed.
> 
>  > I am pretty sure
> > some bugs were fixed on the main itk Imaging classes and you could
> > easily retrieve your direction cosines from your image header (or
> > however the information is stored depending on the image format) and
> > pass it to itk even if the direction cosines are identity
> > 
> > i.e, 
> > this->SetDirection(0, {1, 0, 0});
> > this->SetDirection(1, {0, 1, 0});
> > this->SetDirection(2, {0, 0, 1});
> 
> Well, yes, that's what I do, except that they aren't identity so I set 
> them accordingly.  But when I link, for instance, InsightSNAP against my 
> new library, the program completely disregards the direction information.
> 
> Also, and this is directed more at the developers, I think perhaps the 
> documentation of Get/SetDirection should be changed.  It reads:
> 
> "Get the direction cosines of the image. The direction cosines are 
> vectors that point from one pixel to the next. For ImageBase and Image, 
> the default direction is identity."
> 
> I would suggest something like:
> 
> "Get the direction vectors of the image.  The direction vectors are unit 
> vectors that point from one pixel towards the next.  The components of 
> the direction vectors are the direction cosines.  For ImageBase and 
> Image, the default direction vectors imply the identity transformation."
> 
> This does a better job of keeping scalars, vectors, and matrices 
> straight, and makes it clear to the reader that they should be unit 
> vectors.  Is this right?  Do any developers want to make the change?
>