A new approach for Jawi sub-word segmentation using histogram projection

Segmentation is an important procedure in the character recognition framework which separates one character from the others. Jawi characters as a variance of Arabic characters are written cursively. Segmentation of characters that are written cursively is the most challenging task because both printed and handwritten Jawi characters are written cursively. In this paper, we propose a new approach of Jawi sub-word segmentation. Histogram projection was used to separate sub-words in the segmentation process. Neighbor tracing was also applied to resolve overlapping characters. Generally, space between a word is longer than for sub-word. By using this technique, segmentation from line to word was not necessarily needed. The result of this technique showed that line to sub-word segmentation was performed with high accuracy for printed Jawi characters. This approach could reduce one step in the word segmentation process. The proposed method achieved an average accuracy of 95.13%.