Making it easier to read scans of records when other side of page bleeds through?

7

1

Sometimes the scans of parish registers and other old records show “bleed-through” from the other side of the page. Is there something I can do to make these images easier to read?

Here are some example images, from the old Warwickshire parish records I referred to in an earlier question. The second image is particularly affected by bleed-through. (link to first image)

enter image description here

enter image description here

Verbeia

Posted 2014-07-27T12:18:50.177

Reputation: 2 978

People might also be interested in http://mathematica.stackexchange.com/a/56140/8, where a better method was proposed by another Mathematica user.

– Verbeia – 2014-07-29T22:47:08.527

See also this question: Techniques for reading poor or illegible images. The "levels" technique described there can help some with issues such as this one.

– RobertShaw – 2014-08-02T22:24:38.207

Answers

7

Yes! This is probably possible in any image manipulation program, but I have found that the Wolfram Mathematica software is a very powerful tool for this sort of work. A home use licence costs $A295

The trick is first, to line them up correctly by cropping the right number of pixels from each side, using repeated applications of the ImageCrop function. I haven't found a way other than trial and error to do this.

To get rid of the bleed through on one page, you want to make a mirror image of the other page, and then remove things that are dark on the flipped page and lighter on the target page, and similarly keep things that are darker on the target page than the flipped page. Here's the flipped page:

page17cropflip = ImageReflect[page17crop, Left]

enter image description here

There are a few different things to try, including to Blur the target page so that the text is a bit darker. Here was the set of commands that worked tolerably well in this case (it looks better at full size):

With[{brightness = -0.3`, contrast = 0, final = 0.903`, 
  highbleed = 0.159`, lowbleed = 0.221`, method = "Cluster"}, 
 Binarize[ImageAdjust[
   ImageClip[
    ImageAdd[Blur[page18crop, 3], 
     ImageClip[page17cropflip, {lowbleed, lowbleed + highbleed}]], {0,
      final}], {contrast, brightness}], Method -> method]]

enter image description here

There's a lot of technical detail here but the main command is ImageAdd, which "adds" the two images and thus tends to lighten and cancel out things that are lighter in the flipped image. ImageClip just adjusts the range of shades of gray that are allowed in each image, so that the bleed-through text fades out. There are probably better combinations of commands, but this worked in this instance, and a slightly simpler version worked for the first original image (i.e. blanking out the second).

Verbeia

Posted 2014-07-27T12:18:50.177

Reputation: 2 978

What should one do with that images of pages can be rotated and the page can be scanned not as rectangle but as trapezoid) – GEORG GAAL – 2017-07-10T11:13:17.923

3

When making the copy originally, use a black background rather than the white copier cover. This reduces the contrast on the opposite side and therefore the bleed through.

Opa

Posted 2014-07-27T12:18:50.177

Reputation: 31

1This is an interesting suggestion, but it assumes that the user is making the copy of the record. Many of the images people would like to process are digital images downloaded from sites like Ancestry, FindMyPast, and FamilySearch. They are often digital scans of microfilm images, which were created by the archives so that the original materials are handled as little as possible. – Jan Murphy – 2014-07-27T20:37:01.760

1These are scans from ancestry.com. In general people do not have access to the original documents to improve the scan. – Verbeia – 2014-07-27T20:39:16.483

Welcome to G&FH SE! Thank you for taking the time to post an answer that may later prove useful to someone who does have the option to re-scan. – PolyGeo – 2014-07-28T05:01:16.847