A voting method for image binarization of text-based documents

Abstract

Binarization is an essential stage of image pre-processing before text recognition. This process improves the performance of the obtained text and decreases the effort required for post-processing to find possible mistakes. The main direction for this work is to improve the binarization step in a Document Image Analysis System to increase the text’s accuracy and combine a series of thresholding algorithms into a voting solution.