Edge detection

1
Edge detection
You can find edges in images by subtracting
adjacent pixel values edges show up where they
are different. Whether this works depends on how
sharp the edges are. Printed letters are very,
very sharp. Most stains are not. More
sophisticated is spatial filtering which is an
exact analog of filtering sounds by frequency.
Letters have high spatial frequency, and most
staining will have low spatial frequency. I
first saw this in work by Tamas Doszkocs at the
National Library of Medicine.
2
Edge detection examples
(Alan Ayckbourn works quite well)
3
Less successful
If you think about it, the cat doesnt have any
sharp edges, except for the claws he has fur
everywhere else, even on his ears (making me
jealous in the winter). The whiskers are
actually the most prominent part of the
right-hand image.
4
Bloodstains
Original spot, page 1 Edge detect,
threshold Bleached (hydrogen peroxide)
5
Next page, worse stain
Note stain over letters, and ballpoint pen
writing on page Edge detection removes almost
all the stain but also the pen writing The bleach
damaged the paper (applied too heavily) but
removed the stain without also erasing the
ballpoint ink.
6
Conclusions
Spatial filtering picks up edges but not broad
areas thus it selects letters. Its a more
sophisticated version of edge detection and might
have done better. Once something is reduced to a
fairly dim level, you can remove it with an
intensity threshold. Color filtering is
difficult blood is not pure red. But I could
have worked harder at that. Hydrogen peroxide was
a lot faster than scanning, let alone
experimenting with software. Of course, photocopy
after bleaching, since the paper is now weaker
(assuming you care).