ocr from a website - how and how difficult?

[Denis Wen]

His,

If anyone can share experience with using OCR software with java to parse webpages.
My prospective task could turn out to be parsing a website where unfortunately some of the important information is displayed as images. Therefore I thought it would be better to seek advice first on good (and possibly free:) OCR software to use and maybe someone's portion of experience.

thanks

[salvin francis]

once you have converted a webpage to an image (Quite abstract for me)

you should probably convert it to grayscale and then make custom filters to increase contrast to a very high value,
the resulting output would be an image thats black and white with text ready for OCR


I do not know of any current apis that support OCR.

[Ulf Dittmer]

I'd start by looking at the various packages that a search for "java ocr" brings up.

[Denis Wen]

Thanks for the replies.

No, i was not going to convert the whole website into an image, there are just some pieces of information on the website not in textual form but as png images, that should be converted to textual form with ocr. Looking at the moment at Tesseract (http://code.google.com/p/tesseract-ocr/) since it's free (unlike java-coded Asprise) and quite simple, though should be called from the command line.