Re: Accurate edge detection?

> I don't have an opinion because I never tried to deconvolve a similar
> image.  However, if the results were of aesthetic im****tance, as in
> the cases I typically deal with, I would be concerned with artifacts
> in the areas surrounding the edges.
>
> The best use of theory is in deciding which experiments to
> perform.

I can only repeat- to get rid of these artifacts is very simple. One
only needs to set an appropriate threshold and cut off all what is
above this threshold. The only reason why I didn't do it was that I
wanted to show restoration technique. What actually can be obtained
using relatively simple methods-linear algebra. I do see at least one
possible way to use this theory- to get precise position of a light
dot at the night photo. It usually is surroundid by halo.
BTW, if I'm just here, I can tell you that what really can decrease
some prog is pagination. For example, if you have a huge image what
cannot be loaded to memory, you may only move along its rows, but if
you try to move along columns it dramatically decreases your prog. My
experience with SSE was slightly disappointing- additional ops what
was a must to load data to SSE registers anl later reload them eat all
what I possible saved using group instructions. Interesting, that when
I'd replaced exp(x) with my own small assembler code, speed was
increased, but when I' done same thing with sin(x) result was just
opposite