Gaussian Scale Space

I'm just getting interested in image processing and was reading up about scale spaces. If anyone is familiar with the discipline I have a question. What's the correct way to go about creating a gaussian scale space? The process of the correct way to apply the gaussian filter to a digital image is where I'm initially having trouble. The digital image is discrete, and the gaussian filter operates on a continuous function. I've read that it's improper to simply sample the gaussian at various points that correspond to the discrete points and perform the convolution discretely. It's been mentioned that the modified Bessel function is the correct discrete analogue to the continuous gaussian. My other alternative is to interpolate the digital image in some fashion (bicubic, etc.) and then apply the gaussian to the result of that. However, once again, I've read that doing that is also improper.

Does anyone know what I'm talking about and if so, how should I go about creating a gaussian scale space from a digital image? Start a New Thread

On Thu, 4 Jun 2009, Donato Richardson wrote:

> I'm just getting interested in image processing and was reading up
> about scale spaces. If anyone is familiar with the discipline I have a
> question. What's the correct way to go about creating a gaussian scale
> space? The process of the correct way to apply the gaussian filter to a
> digital image is where I'm initially having trouble.

http://en.wikipedia.org/wiki/Scale-space_implementation

--
Kind regards,

Filip Rooms

Universiteit Gent - TELIN/IPI/IBBT
Sint-Pietersnieuwstraat 41, 9000 Gent
+32 9 264 34 15

http://telin.ugent.be/~frooms/

"...very few phenomena can pull someone out of Deep Hack Mode, with two
noted exceptions: being struck by lightning, or worse, your *computer*
being struck by lightning."
(By Matt Welsh) Start a New Thread