I am looking for a way to obtain the position of individual characters in the OCR output. I can get the position of each word, but not of the individual character. Is there a way to do this?
I considered looking at the strings and then dividing them, either into equal parts or into parts based on the “width-profile” of the character.
However, for Arabic that approach will not work as the language contains text “extensions” (tatwheel) which may be placed between characters to create a more aesthetically pleasing text.
Tricky issue. Your “divide string length by number of characters” workaround is the only solution I know.
Also, I am not aware of any other OCR API that gives you the individual characters. Some do not even provide the bounding box of words (like ocrspace does), but only return the xy of complete lines.
Hi Plankton,
Several OCR solutions (incl. in API version) deliver character-level coordinates, incl. Abbyy and Google OCR as the most well-known. Clearly, these are more costly solutions. However, the interesting part here is that (from what I have seen), OCR space is actually on par with these solutions when it comes to character recognition! That is why - from my perspective - it would be very, very interesting if OCR Space could deliver character-level coordinates.
Secondly, I would be surprised if these coordinates were not available internally, although, of course, I do not know how OCR Space derives the characters
