A Thresholding Algorithm for Text/Background Segmentation in Degraded Handwritten Jawi Documents

Abstract

The old documents in Jawi script are being used widely for references. The hard copies of those scripts will deteriorate as time passes. Most of the previous works on Jawi documents focused on the character recognition and the accuracy of the algorithm was very much affected by noise. An effective preprocessing method is required to binarize degraded Jawi document. In this paper, a new technique to threshold degraded Jawi document is proposed. The results of the new algorithm were also evaluated and compared with several algorithms. The quality of the thresholding methods was assessed using visual inspection and mathematical evaluation. The results show that the new technique has outclassed other binarization algorithms.