The Model and Implementation of Javanese Script Image Transliteration

摘要

Manuscript image transliteration system is an implementation of transliteration model of manuscript in Javanese script, which is an alternative for transliteration issues in terms of length of process and accuracy of transliteration result. One of the important stages in manuscript image transliteration system is transliteration of every script image in the manuscript image. The main problem of transliterating Javanese script image of manuscript is high variability of script images in terms of size, thickness, and slant. The main purpose of this research is to find the value of accurate recognition of Javanese handwriting script by applying some model in preprocessing process and feature extraction. The template matching approach is applied in the process of labeling a script according to its class. Since handwriting characteristics are most likely not the same in how to write a script, in preprocessing processes are made to uniform character properties in terms of color, thickness, slant, and size. Since the Javanese script similar with the Thai script, the characteristic search approach applied to recognize the Thai script is also applied in this research. The main data for test data source was the result of segmentation of manuscript photocopy digital image with catalogue number SB.141 which is stored in Sonobudoyo Museum has mbata sarimbag writing style. The result of testing script image transliteration system on 291 Javanese script images produced 95% confidence level, showing that the confidence interval of the average percentage of truth of the transliteration is 73.51% to 85.69%.