QR code cleaning after scan

[sucotronic]

I've developed a script to process scanned documents that carry a QR code on them. When the original is scanned, I apply median 2 and noise 2 and the QR decoding library (zxing more used for android), and normally it decodes ok. The problem is when somebody scan a not original document. I've tried to vary values of median an noise, but the qr library is not able to indentify it. Any tips in additional parameter to get better results?

Here are my tests so far:

[fmw42]

try using -threshold and -morphology

convert test.png -threshold 75% -morphology open square:1 result.png

see
http://www.imagemagick.org/Usage/morphology/


Or try multiple iterations of -despeckle (and possible threshold afterwards)

convert test.png -despeckle -despeckle -despeckle -despeckle -despeckle -despeckle -despeckle -despeckle -despeckle result.png

see
http://www.imagemagick.org/script/comma ... #despeckle


If on linux/mac osx or windows with Cygwin, for automatic thresholding try my scripts, otsuthresh or kmeansthresh or isodatathresh. See my link below.

[sucotronic]

try using -threshold and -morphology

convert test.png -threshold 75% -morphology open square:1 result.png

see
http://www.imagemagick.org/Usage/morphology/


Or try multiple iterations of -despeckle (and possible threshold afterwards)

convert test.png -despeckle -despeckle -despeckle -despeckle -despeckle -despeckle -despeckle -despeckle -despeckle result.png

see
http://www.imagemagick.org/script/comma ... #despeckle


If on linux/mac osx or windows with Cygwin, for automatic thresholding try my scripts, otsuthresh or kmeansthresh or isodatathresh. See my link below.

Thanks! I also have to add -scale 30% because it seems that the QR decoding engine works better with small images instead of full 300dpi A4 scans