Re: [APML] Scanner streaks

[a9125657@unet.univie.ac.at]

Hi Michael, hi to the group,

Basically a good and well maintained scanner and should not generate a
pattern, but mine (a CanoScan 2700F) does it as well when I scan slides;
it looks like woven texture, just as Michael describe it. Ok, time to
reveal one of my image processing secrets:

First I tried to establish if the pattern is constant. Had it been so, it
would have been quite easy to correct for it: Just scan an unexposed
(black) frame and use it as bias frame, i.e. subtract is from all picture
frames, just like the CCD people do it. To reduce the additional noise
introduced by the subtraction of two images, it is of advantage to average
more than one unexposed frame to get a smoothened bias frame. 

Unfortunately, I found my scanner bands not to be constant, but to vary
from scan to scan. But they were always exactly aligned in horizontal and
vertical direction. So I could use another trick to gereate the desired
bias frame: When scanning an image, I also scan part of the unexposed
image borders. To correct for vertical banding, I first select the upper
or lower unexposed horizontal border and define it as pattern. Then I
create a new image with the same dimensions as the picture frame and fill
it completely with the pattern. It will be filled with vertical lines. I
then smooth it only in vertical direction as I do not want to introduce
additional scanner noise to the picture frame. The image thus generated is
a bias frame for the vertical scanner streaks, to be subtract from the
picture frame. To correct for the horizontal scanner pattern, I repeat the
same process using the unexposed vertical border to generate the new bias
frame.

In some images, I found the scanner pattern to be more multiplicative than
additive. In these cases I divided by the bias frame(s) instead of
subtrating. All this can be done in Photoshop.

Be sure to mask the parts of the image overexposed during the scanning
process (stars etc.), as they are not subject to the scanner pattern but
would be so after bias correction.

To get rid of slight residual bending in already corrected images, I like
to apply a mild "dust & scatches" filter only to the background. 

I hope that helps, and I wished some other APML experts would also reveal
a few of their image processing secrets to the list,

  Walter Koprolin
  Vienna/Austria/Europe
  Gallery of Astrophotography: http://www.astro.univie.ac.at/~koprolin/

> Here's a very basic scanner question. 
> 
> My scanner (Minolta QuickScan 35 Plus) generates a cloth-like "woven" 
> texture of faint lines when I scan a slide that is very dark and turn up
> the levels so that the background is not black. 
> 
> I presume these are scanner artifacts and all scanners generate them.  True?
> Or do I have a bad scanner?
> 
> How do you deal with them?   Duplicate the slide onto negative film so it
> will have less dynamic range?
> 
> Can you trick the scanner by telling it it is scanning a negative, then
> inverting the image later?
> 
> 
> Clear skies,
> 
> Michael A. Covington  /  AI Center  /  The University of Georgia
> Author, ASTROPHOTOGRAPHY FOR THE AMATEUR
> http://www.CovingtonInnovations.com/astro  <><


--  APML Archives at <http://www.system.missouri.edu/apml/>  ---
                Unsubscribe at <majordomo@seds.org>