SOT: Scanning old computer docs

[Guy Dunphy]

Hi Kevin,

Are you _sure_ the two modes use the same optical sensor?
Often sheet feed systems use a different sensor somewhere along the paper path.
I suspect they are not, given the odd colour pattering off the top edge of the
paper, on the autofeed image.

If those two images are definitely produced by the same sensor, through the
same glass, then there are a few other possibilities:

* A 'mode settings' problem. Scanners always have a bunch of modes, such as
  photo vs printed, sensitivity adjustment curve, gamma correction, etc.
  It may be your machine uses different settings for flatbed vs page feeder,
  and the page feeder settings are _way_off_.

* Keeping the paper flat on the glass. Someone else pointed this out - maybe
  the sheet feed process uses rollers and the paper is bowing away from the
  glass in those two really bad wide streaks. But there are other problems too,
  because there are fine streaks as well.

* Another thing scanners do, is calibrate the sensor using a standard white/black
  image bar, usually behind one end of the flatbed bezel.
  The sensor always has imperfections, but subtracting the observed calibration
  data results in a clean image. (Ideally, sometimes not.)
  Maybe your machine uses separate calibration datasets in the two modes, and the
  one for autofeed is rooted. Download a manual and read.

There are defects in your 'good' flatbed image too - for eg the bleed-through
of the orange lettering on the other side of the sheet. The way to correct that
is to use a black, highly light absorbent backing sheet. Eg black velvet.
But you probably can't do that on your scanner. This defect is typical of scans
done by people with little discernment for image quality. They don't even notice.

Then there are fundamental defects common to both images:

* You've used PDF as the end format. As a result the image is stored in lossy JPG coding,
  and thus has typical JPG edge noise. PDF does not allow any sensible image coding
  scheme such as PNG. Just a few old ones: JPG (OK for photos, but awful for any image
  with fine detail and hard edges), TIFF (lossless but very inefficient), the horrible
  'Fax mode' (B&W, in which the captured data is stored losslessly, but the capture
  process is horrifically lossy and destructive of image quality.)
  Plus another abortion called JBIG2. ( A patch tokenization scheme that is so bad it
  should be illegal.)

* The originals are screened printing, where all tones are created by varying the size
  of saturated YMCB ink dots laid in a regular grid. To scan this kind of document
  adequately, you HAVE to scan at high enough resolution to resolve the edges of the
  tiny dots (the only way to avoid obnoxious moire patterning), then post-process the
  image to convert (blur) the screen grid dots to smooth colour shading (still at very
  high resolution) then scale the image to a desired final size. Then lastly change the
  image encoding to a suitable (lossless) form (eg one of the PNG modes) for sane file size.
  This is still unavoidably a manual page-by-page process if you want archival quality.

  And PDF flatly doesn't allow it. Once you realize this is so, you'll become very
  depressed about all the people who think they are 'saving worthy documents for posterity'
  by scanning them into PDF.

  Your images still contain remnants of the screening dots pattern, but messed up
  by JPG artefacting. Thus the file size is stupidly high due to all the superfluous
  and messy fine detail.
  Zoom right in to see it.

Sorry to not be very helpfull,
Guy



At 11:56 AM 3/04/2021 +1100, you wrote:
>If there is anyone on the list familiar with scanners I'd be most grateful
>for some advice please.
>
>Some time ago I bought a HP 8270 sheet feed (full duplex) scanner NOS. I
>wanted to digitise a whole heap of old computer documentation and for a
>little while I've been working through the big heap of stuff. But for quite
>some time I've had an issue with scans that go through the sheet feeder
>(irrespective of whether I do them double sided or not). Basically the
>problem is that anything that goes via the sheet feeder has issues with
>"streaks" in the document whereas anything done on the flat bed is perfect
>(I have some links to some examples below).
>
>By way of clarity, anything done on the flatbed the lamp traverses the flat
>bed to do the scan. For sheet fed items the lamp is moved to specific slot
>on the scanner and the sheet feeder takes over wrapping the document past
>the lamp. Given that flat bed scans are OK I don't think its an issue with
>the lamp.
>
>I've done the following things to try to resolve the issue with no joy:
>
>* Checked for any specific settings
>* Tried doing scans in grayscale
>* Tried increasing the resolution (default is 300dpi) to slow the speed that
>the document is fed through the feeder.
>
>According to HP the issue is a cleanliness one i.e. dirt on the glass can
>cause reflections. I've followed their instructions for cleaning the glass
>but still no joy.
>
>Flatbed  example.
>
>https://drive.google.com/file/d/1jYiFzERiZiaq7-WoTiQ2eIzITn6giviR/view?usp=s
>haring
>
>Sheet feed example.
>
>https://drive.google.com/file/d/1WF4SHbwV3bVET_bzwIUiULywbCGZBf_V/view?usp=s
>haring
>
>Thank you!!!
>
>
>
>Kevin Parker
>
>
>