please give me an idea for enhance the left image ( image 1 )like (image 2), I tried contrast enancement algorithms (HE,BBHE,M​MBEBHE,CLA​HE,RSIHE,.​.) but i didnt get satisfied responsese. Do you have any idea for binarize images which have shadow ??

please give me an idea for enhance the left image ( image 1 )like (image 2), I tried contrast enancement algorithms (HE,BBHE,MMBEBHE,CLAHE,RSIHE,..) but i didnt get satisfied responsese. Do you have any idea for binarize images which have shadow ??

Answer 1

[Image Analyst]

Try using adapthisteq() on it.

[Image Analyst]

I could give you an algorithm that works for that one, specific image but then you'd just come along with another image and say "But it doesn't work for this image." So what you need to do is to look up the OCR algorithms that people have been perfecting over the last 30 years or more. I bet they'd be better than any amateurish thing I'd suggest. They're all listed here: http://www.visionbib.com/bibliography/contentschar.html#OCR,%20Document%20Analysis%20and%20Character%20Recognition%20Systems So pick one and code it up. I know it's work and not going to be some 20 line solution but if you put in the effort you'll have something that works better.

I've attached a document thresholding demo that uses an algorithm I saw at a conference a year ago.

[Image Analyst]

I'm reluctant to toss out any more ideas because you've apparently ignored the first three I suggested (VisionBib, adapthisteq, and my m-file). At least that's how it appears. If I suggest another or code up something, you may ignore that too. I'm not an OCR expert. But I'll give you one more. The University of Dayton has done work in shadow removal: https://www.udayton.edu/engineering/centers/vision_lab/wide_area_surveillance/visibility_improvements.php Contact them for help applying it to your images.