Implementasi Projection Profile dan Connected Component untuk Segmentasi Citra Manuskrip Beraksara Jawa Cetak

Abstract

Javanese script is one of Indonesia's cultural heritages. Many ancient manuscripts in Javanese script are still neatly stored in museums and libraries in Indonesia, but only a few people can utilize the important information contained therein. The difficulty of the transliteration process is one of the challenges in obtaining this important information. With the development of document image analysis science, this research was developed to help shorten the process of transliterating Javanese manuscript images. Segmentation is an essential stage in transliterating the manuscript image, namely, automatically taking each script image in a document. This research developed the segmentation process by combining the projection profile and connected component methods. Using one image from a scanned manuscript in the book "Hamong Tani" written using printed Javanese script on page 5, the results of line segmentation with 100% accuracy and the results of Javanese script segmentation with 95.952% accuracy were obtained after preprocessing. From the large segmentation accuracy value, it can be concluded that the projection profile and connected component methods can be used well in segmenting printed Javanese manuscript images.