Unfortunately, this is not CSI, we collected 20,000 images for license plates; we were wondering if there is a semi reliable way to read those license plates using OCR. the images especially night time ones are extremely low quality.

  • 2
    What, exactly, are you doing? This is either law-enforcement, a student project, or reeeeeealy creepy. – Fake Name Jul 9 '10 at 7:12

Tesseract is the best (free) OCR software that I've used. I've used it for scanned paper documents but it might also work for license plates.

You would need to convert all the images into TIFFs, if they aren't in that format already.

| improve this answer | |

Most off-the-shelf OCR packages are designed to recognise text from scanned documents or similar and expect reasonable text alignment, angle of view and contrast settings.

You will probably need to do pre-processing on your images to (a) detect and crop to the area of the image containing the license plate, (b) perform perspective correction and optionally (c) contrast and noise enhancements for the night shots.

| improve this answer | |

You might have better results with the images if you pre-process them to remove noise using something like this tool:

Noise Ninja at PictureCode

I have found references to open source plugins for the popular GIMP tool if a commercial solution isn't acceptable to you, but I've never been a big fan of that tool (GIMP) personally.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.