How can I 'subtract' an image on one side of scanned page from the other side?

In Acrobat Pro....

if you have a PDF document scanned, you can substract or cut any page you want.Still, not so sure that solution works for you

philstubbington said:

I am trying to transcribe a diary my grandfather wrote whilst he was in a prisoner of war camp in the Second World War, and unfortunately the paper is so thin in places that the handwriting from one side of the paper is visible on the other side. This makes it impossible to read the text on either side.

Is there any software that would allow me to take a scanned image of both sides of the paper and separate out what was originally written on one side from the other? I don't have any use (or expertise) in any graphics software so something very easy to use would be ideal!

Would be grateful for any advice.

Thanks.

Click to expand...

Experiment by placing your document over various background pieces of paper to minimize the bleed through... and then take a high resolution photograph.

Today's cameras will do as good a job as a scanner in many cases. Make sure that the "film plane" (actually "sensor plane") is parallel to the document.

/Jim

I agree with trying different paper behind the pages when you scan. Many scanners will use white to maximise brightness and reflectivity, but you might be better with a darker piece of paper. Experiment.

Once you have the scans, you could try to edit the images (in any image editor, from iPhoto to Pixelmator to Photoshop / Elements) and use the contrast and brightness controls to increase the contrast as much as you can, and perhaps slightly increase the brightness.

Depending what app you use there are also tools like 'threshold' and 'posterise' (try a low number of levels) that may also reduce the writing from the other side of the paper.

What you are essentially trying to do is make use of the fact that the writing on the side you want should be ever so slightly darker and bolder than the writing on the other side, and make the computer narrow down those slightly darker and bolder pixels from the slightly lighter and less bold pixels.

I hope something in all that helps, good luck.

Try popping a sheet of black paper behind the sheet you are scanning. This, in theory, should merge with the dark ink on that 'back' side of the sheet. Then with some exposure and contrast adjustments the writing on the front should be legible.

Good Luck. If you need more help, just post back... I think you will find that there are more than a few of us who would like you see some success with this project.

Let me see if I understand.

You don't actually have the diary in hand.
What you do have are photocopies of each page.
When the photocopies were made, it picked up the text from the back of diary page, along with the text from the front.
So, in effect you want to work with someone else's scan? (If we call the initial photocopy a 'scan').

Is that an accurate summary?

If so... that makes life more interesting.

Except.... you will note on the examples you provided, there is black text and grey text. If you play with the levels/curves you can get rid of the grey text. With any luck the majority of the grey text will be the stuff from the back, and not the front.

If there is a lot of grey text on the front (faded ink, etc) then of course it that will disappear as well... not a good thing. Hard to tell from the small samples you posted.

snberk103 said:

Let me see if I understand.

You don't actually have the diary in hand.
What you do have are photocopies of each page.
When the photocopies were made, it picked up the text from the back of diary page, along with the text from the front.
So, in effect you want to work with someone else's scan? (If we call the initial photocopy a 'scan').

Is that an accurate summary?

If so... that makes life more interesting.

Except.... you will note on the examples you provided, there is black text and grey text. If you play with the levels/curves you can get rid of the grey text. With any luck the majority of the grey text will be the stuff from the back, and not the front.

If there is a lot of grey text on the front (faded ink, etc) then of course it that will disappear as well... not a good thing. Hard to tell from the small samples you posted.

Click to expand...

I agree... the issue is that your copy actually has all of the data on the same side... not being bled through from the back. Hence... or previous advice would not work.

Your best success would be to pull this into a photo manipulation program and play with levels/curves and contrast to "erase" the lighter text that was originally allowed to bleed through on the initial (poor) scan.

Is there any chance that you can get access to the original again? Scanning it with more care would give the best results.

/Jim

philstubbington said:

Thanks everyone for your help so far.

....

Click to expand...

Can you post a bigger version of the scan? Or a link perhaps? Just one page... I tried playing around with levels and curves, but the text is so pixellated in the images you attached I can't tell if it's cleaning up or not.

Cheers

philstubbington said:

Here's full page samples.

Good luck!

Click to expand...

I spent just a couple of minutes using the "curves" function of Aperture 3. I am not too pleased with the results... but with care... you can read most of the text (difficultly).

It would be really great if you had access to the documents to copy again... this time with more care. I would personally photograph the pages with a dark sheet between pages.

Another option is to post this request on the "retouching" forum of DPReview. Here is a link to the actual forum: http://forums.dpreview.com/forums/1006

If you decide to post there... include "for play" as that is a keyword people seem to use to try their skills on various projects. For example... you might title your post as "Journal Pages with bleed through - for play".

Also... if you do get good results from the experts over on DPReview.com... please post the results her so we can see how well you did.

/Jim