Thanks to those who answered

llya

Yes I would be interested. Paint Shop Pro calls it "Moire Pattern Removing"
which I found very useful especially when I wanted a photograph from a
newspaper. At the present time I can use "Descreen" on my new printer/scanner
to remove the Moire pattern for new pictures but that is no use for some of my
older images.

I'm using Windows XP

On Fri, 22 Feb 2008 22:25:31 +0300, shampayne
<webforumsuser@macromedia.com> wrote:

> Thanks to those who answered
>
> llya
>
> Yes I would be interested. Paint Shop Pro calls it "Moire Pattern
> Removing"
> which I found very useful especially when I wanted a photograph from a
> newspaper. At the present time I can use "Descreen" on my new
> printer/scanner
> to remove the Moire pattern for new pictures but that is no use for some
> of my
> older images.
>
> I'm using Windows XP
>

Oh yes. PSP is quite cool when it comes to bitmap editing, but can not
compete with FW when it comes to vectors, and vice versa.

Anyway, the filters in both programs are built to be "user friendly", that
somewhat limit their capablilies.

I'll try to explain the technique I personally found to be most efficient.

First, download the freeware "Image Analyzer" from

http://meesoft.logicnet.dk/

The key feature you need is called "Frequency domain filter".

Now a bit of theory on Fourier transform and frequency spectrum stuff.

FT is about representing the image (or other data) as a composition of
sine functions (sine function is periodic or "harmonic" wave). From
ordinary average guy's point of view, describing the photo as huge set of
waves added together is senseless, but scientists think different.
Interesting thing happen, if our image *really* have some periodic
component in it. Regardless of what is this component - periodic pattern
of printed halftone in image, or result of interaction of periodic
halftone with periodic nature of scanning which we know as "moire" - it
will give us high peak in frequency spectrum (that is, in image
representing how intensive are waves of dirrerent frequencies). We can
detect that peak and, if necessary, edit or even remove it, thus editing
or removing this particular signal from the image.

Now we go to the practical part. You should scan your images at ppi at
least twice as high as printing halftone screen lpi (remember of Nyquist
theorem). Most printings are done with lpi about 150, and newspapers often
at even lower values, so you should scan them at 300 ppi at least (and
without any software sharpening scanner may offer you). Then open image
with Image Analyzer. Then go to Filter -> Frequency domain filter. Now you
see the frequency spectrum of the image. In the dialog, you may need to
click "View scale" down scale a couple of times so the image is easier to
view. The image you see is likely to look like the galaxy of white stars
spreading from the top center part to the edges. The top center part is
brightest. Beside it, you are likely to see some additional bright stars
(high peaks) apart from it, forming some sort of a pattern. These are the
peaks representing our "parasite" periodic pattern in the image. So you
should just paint black over them, checking out the intermediate result
with "Preview" button. Important point is: don't hit the bright spot at
the top center area, this is useful signal. Only hit the "lone stars" far
from it. Once you're satisfied, hit "Ok" and save the image.

Althought the technique may look complicated, I often found it to be most
efficient when dealing with halftone or moire.

--
Ilya Razmanov
http://photoshop.msk.ru - Photoshop plug-in filters