[Insight-developers] DifferenceOfGaussiansGradientImageFilter

Damion Shelton beowulf@cs.cmu.edu
Wed, 22 Jan 2003 17:33:58 -0500


Hi,

This class does gradient computation prior to running the core atom=20
code. There's nothing particularly special about what it does.
> =A0
> 1) this code is inefficient. the same calculation is repeated done in=20=

> an inner loop

I'll take a look - this was written back in the spring of 2001, so I=20
was rather inexperienced with ITK at that point.

> 2) the filter is is not handling no properly managing the input=20
> requested region

Yup, that's true.

> 3) the calculation does not appear to be a difference of gaussian=20
> gradient calculation

I'll defer this to George - it's based on his code. As I understand it,=20=

this filter approximates a DOG computation when run in tandem with the=20=

BinomialBlurImageFilter.

> =A0
> What is the status of this filter? Is it under development?
> =A0
To be honest, this was one of those "temporary fixes" that worked well=20=

enough to not demand immediate replacement. It could easily be replaced=20=

by any similar class that does a decent job of computing gradients (one=20=

did not exist at the time I wrote the code).

-Damion-