A Local-to-Global Approach for Document Image Binarization

Abstract

Document image binarization is a preprocessing step in Optical Character Recognition (OCR). This paper presents a novel hybrid approach for degraded document image binarization. An adaptive or local approach is applied to enhance the degraded image that separates the foreground and background considerably. After that, a new global approach for binarization is proposed. A postprocessing technique is applied after to get the final output. This transition from local-to-global approach results in document image binarization. Exhaustive experimental analysis has been carried out on a series of benchmark dataset having variations of degraded document images. Comparative analysis of the proposed method with other existing methods promises the effectiveness of the approach.