[Insight-developers] Appropriateness of filter data types

[Joshua Cates]

Hi,


On Wed, 4 Apr 2001, Luis Ibanez wrote:

> Maybe the use of NumericTraits could provide this functionality,
> something like:
> 
>    float   NumericTraits<myRealType>::GetFloat();
>    double  NumericTraits<myRealType>::GetDouble();
> 

Do you mean using traits to automatically convert to floats during data
access within the algorithm? Yes this is one option and may be a better
way for sparse access.  I suppose that data accessors could even be used
in this way to convert input to floats.  But for dense access, the
copy-to-float strategy requires at most one conversion per pixel, a speed
advantage at the expense of memory.

I'll work up a sample of what I'm talking about and post it somewhere
for interested folks to look at.

Josh.