I'm facing huge trouble recognizing the characters as in the pictures below. I've tried basic preprocessing like increasing dpi, grayscale, crop, deskew but glare in the characters due to light reflection appears to be the problem. Images are unordered vehicle number plates. I'm using tesseract 3.05 on windows 10 and tessedit_char_whitelist is limited to A-Z and 0-9. Any suggestion, is appreciated.
Thanks.





