Ask HN: Recommended OCR?

9 pointsby shiny16 anni fa8 comments

I'm OCRing a bunch of TIFF files with tesseract, and while it works to some degree, it's nowhere near as accurate as I'd like it to be. Perhaps I'm doing something wrong and I could tune it to my liking, but I can't find too many resources on tesseract. Am I missing something?

Any other recommendations for OCRs? Ideally it would be free, but I'm willing to pay if it's not too pricey.

I've been trying out the trial version of FineReader, and it seems to work pretty well, so I may go with that.

Any help is greatly appreciated.

8 comments

MaxGabriel16 anni fa

I've had really great success with finereader. I tried out every free OCR tool I could find and after poor results went for finereader.

Spend some time on their website so you get the right product, they have multiple prices for the same products, too. I got the latest Finereader (after a coupon code I found on google) for between 130-150.

(I'm mostly scanning books)

ig116 anni fa

Finereader is what Project Gutenberg has been using for the last decade or so.

hebz0rl16 anni fa

what about gocr? its opensource see http://jocr.sourceforge.net/

hijimayor16 anni fa

tesseract is way better than gocr, and is also open source

usermac16 anni fa

Fujitsu ScanSnap

hijimayor16 anni fa

One thing that improves Tesseract's performance dramatically is giving it grayscale tif images. Do

mogrify -type Grayscale *.tif

and run them through tesseract to see the difference. No idea why no one mentions this in the documentation.

mgedmin16 anni fa

I've found that Tesseract's accuracy was much improved when I converted my grayscale images to bitmaps with GIMP's Threshold tool.

(Those pictures were not-very-hight quality snaps of a few pages of a book taken with a 3.2 Mpix digital camera.)

hijimayor16 anni fa

Tesseract really wasn't made for camera pics, but it could use some help with the thresholding :)