Scanning hints (Wouter's BBC pages)

This what I do to get really good quality scans (note that I'm not interested in OCR'ing, as all the programs I tried usually produce complete and utter junk in almost all cases):


Before you start

You need to make:

  1. a cardboard sheet, with one side completely black.
  2. a piece of paper also with one side black; both sides black is even better.

Both must be at least the same size as the pages you want to scan, usually A4.


Choose scan resolution

This depends on the scanner and image that's being scanned. Depending on the scanner and the dot resolution that the original image was printed at, you can get artifacts/patterns. I usually scan at 300 dpi, then reduce image sizes as required as at 150 dpi this often happens. Lower resolution scanners will usually have less of a problem scanning at lower resolutions.


Scanning

Put the cardboard sheet on the reverse side of the page to be scanned with the black side down (facing that page), or if this is not feasable (e.g. for magazines), use the black paper.

The black side of the cardboard/paper will absorb the light going through the page to be scanned, and keep it from reflecting back which would give a strong ghost image of the reverse page.

By using the cardboard & heavy object on top of that, you will get rid of most wrinkles.

You can also use sticky tape to fix the black paper to the lid of the scanner so all images get scanned in this way, and you can still use the cardboard if there's a need... Strange that all scanners have white backgrounds. Images will get brighter but at the huge cost of strong ghost images.


Image manipulation

Now you get to the image processing/enahncing stage: For greyscale images you may want to smooth the images by shifting the intensity of images (dark -> darker, light -> lighter, see the image from 'xv') to get rid of almost all wrinkles and ghost image residue. You'll need to fiddle with your image processing programs to get the best results...

This also reduces file sizes a lot. The downside is that this removes the texture from the paper; as that's not usually of interest, this is a good thing.

After that , reducing the number of colours (gray levels) to about 8 works very well in most cases (i.e. reduces file sizes a lot and usually doesn't hurt image quality). You can use Image magick using the command:

convert -colors 8 <input file> <output file>

to convert a file from e.g. tif to png and reduce the gray levels.

I used to enhance contrast with xv and later used my own scan processing programs. One particular problem I face with my scanner is that this gives darker/fuzzier sections where the page to be scanned is slightly away from the scanning plate (happens e.g. with paper that has been folded, edges of magazines, and paper that has been damp/wet... See the ice-box example scan on this page). My new enhancer takes care of most of this problem for b&w images, but for full colour it's still a big problem. Anyone have a good and fast A3 scanner for me?


Example images

1. Standard scan:

Note the wrinkle (middle bottom) and the ghost image of the reverse side of the page...

2. Scan after placing a black piece of paper behind the page to be scanned, and then placing a book on the magazine:

Note the much less visible wrinkle (middle bottom), and the ghost image is almost gone (yes, it's still there, compare closely to the first image, and you'll be able to see it).

3. Previous scanned image edited by making light->lighter, dark->darker:

Note the hardly visible wrinkle (middle bottom) and the ghost image seems gone.

To email me, go to this page