OCR voting methods for recognizing low contrast printed documents

Modern adaptive thresholding algorithms do their best to provide good quality binarized images. Unfortunately, it\´s hard to find a good compromise between the amount of background noise in the binary result and the amount of breaks or missing parts in the shape of characters if the original grey image has low contrast. In this paper, we describe some voting methods starting from an external "black box" voter, to a more deeply integrated "shape" voter that can be used to generate even better recognition results by running a voting OCR engine on two, differently thresholded images