ocr and image preprocessing techniques

Question

Hi all I have a collection of images that are similar to photos of car registration plates.

I am wondering how to pre process the image before sending it to an OCR engine to determine the text.

I would like to eventually be able to crop the input on only the white content of the registration plate.

The actual image would look something like this:

and should end up something like this (color depth not important)

Thank for any help. Regards, Bob

Answer 1

• Segment the image (take a look at this).

• Classify segments using a gray scale threshold (since you're just looking for white).

• Crop the image by using the white segment bounding box (or bounding boxes, if it breaks down in more than one segment). This is probably not going to work for white cars - but it should not be too difficult to establish a few rules for this case.

• Take a look at tesseract-ocr to extract the text.