Re: Accurate edge detection?

illywhacker:
> On Jul 29, 11:55=EF=BF=BDam, Andrew_M <m...@[EMAIL PROTECTED]
> wrote:
> > BTW, when I was working with restoring of diffracted images, I quickly
> > realized, that interferention is in fact nonlinear effect. You
> > actually measure only intensity of a vawe field, but sum of even two
> > vectors doesn't meant, that lenght of a result will be equal to sum of
> > lenghts of components.
>
> This is true.
>
> > But this is the nice sample of the fact, that complicate problems
> > cannot be solved, using simple methods. Each method must be adequate
> > to problem to be solved.
>
> I agree. And this is a complicated problem. Although maybe you do not
> have to deconvolve the whole image to estimate the edge position. If
> you know you are dealing with a straight knife edge, it is a one-
> dimensional problem if all other parameters are known.
>
> illywhacker;

Due to the fact that Mr. Spencer didn't explained in details what sort
of objects he is working with, I have a guess that =C2=ABedge knife=C2=BB
i=
s the
term what's only purpose is to make further discussion simpler.
Probably he wants to get better resolution than his optics plus
scanner allow to get. This is possible sometimes for some sort of
objects, but with much more complicated methods than at what steepness
of a slope true edge is?. Really, when I had made my images, I found
one old book on computer application in optics research, where I found
description of a methods, likes what I actually used to restore images
after diffraction. At this book some research works are described, but
really old ones- made 30 years ago. But classic methods probably still
functional. Yeah, this is non-linear problem, as I wrote. I did right
things, as you may notice, trying to compare two images:
plight://www.smartfills.com/Html/Images/diffr1.jpg (the first
iteration) and  http://www.smartfills.com/Html/Images/diffr3.jpg
(the
eight iteration). After this iteration the source could be restored
using simple threshold. In any event,.contours was restored more than
accurately after the first iteration. Therefore, for two colors images
this method solve the problem completely. The only thing, what isn't
clear, is why Mr. Spencer asks for advices and after it bravery
declines them all.