Adaptive Thresholding for Document Image Binarization
Binarization is the process of converting a colour or grayscale image into a binary (i.e. black and white) image. It’s typically the first step performed by software that does any sort of document analysis and recognition, optical character recognition (OCR) being the most common application. When a document is scanned with a flatbed scanner, well established algorithms do a very good job. Nowadays, however, many document images come from the cameras in our mobile devices, and these are subject to variable lighting conditions that make binarization quite difficult. The problem is exacerbated…