A simple and novel adaptive binarization approach for handwritten documents

Handwritten text recognition needs some preprocessing steps for better recognition. One of the preprocessing task is handwritten text document binarization. Several binarization approaches are proposed previously and several are widely used in handwritten text binarization but the choice of most appropriate binarization approach for handwritten document is a very difficult itself. In this paper, we propose an adaptive binarization approach which can handle both continuous and abrupt intensity variations across the lines as well as words for handwritten or printed document. Experiments found that, the proposed approach is very simple as it uses only addition operations with constant complexity, and it also gives competitive results for handwritten (printed) documents compared to standard binarization approaches.