Penemberengan Teks Jawi Tulisan Tangan: Satu Pendekatan Gabungan

This article explains a combination approach of segmenting Jawi text. Segmentation is one of several main functions in Jawi Optical Character Recognition or lOCR. It involves a process of separating a collection of text to characters for recognition. In general, the text have five basic forms which are; vertical overlap, ligature, diacritics, horizontal overlap and two connected characters. There are three main approaches to segment these forms, there are Histogram Profile Projection (HPP), Labelled Connected Components (Lee), and Determining of Segmentation Points (DSP). HPP can be used for segmenting Jawi text to text lines, then to words. Lee can gather all contours of connected components, meanwhile DSP is stressed on determination of definitive segmentation points by searching the junction segments between characters. These three approaches are combined in order to solve the problem of segmentation for Jawi handwritten text with a little modification. The related algorithm is also described which emphasises on three main forms ofJawi characters; which are vertical overlap, ligature, and horizontal overlap. An experiment has been carried out and the results are discussed in comparison to those of HPP approach.