RESTORATION OF OLD MALAY JAWI MANUSCRIPTS USING MUMFORD-SHAH AND BERTALMIO INPAINTING MODELS

Abstract

Jawi is the earliest writing script of the Malay Archipelago that was derived from Arabic letters. Old Malay Jawi manuscripts may provide vital information about the legacy, cultures, and historical evolution of the Malay Archipelago over time. However, while preserved, old Malay Jawi manuscripts tend to be damaged. Image inpainting is a process of reconstructing missing parts of an image which can be used to restore the old Malay Jawi manuscripts. The Mumford-Shah and Bertalmio inpainting models are two well-known and effective methods for solving the image inpainting problem. Hence, the aim of this study is to determine which model is better at restoring corrupted input images of the old Malay Jawi manuscripts. A sample of thirty (30) old Malay Jawi manuscript images were obtained from Kumpulan Penyelidikan Etnomatematik Melayu (KUPELEMA). The corrupted images were restored using both models, implemented using the MATLAB software. The Structural Similarity Index Measure (SSIM) and Mean Absolute Error (MAE) were utilized to assess the quality of the results. The numerical experiment demonstrates that the average values of SSIM and MAE for Mumford-Shah inpainting model are 0.9380 and 0.0151 respectively, while the values for the Bertalmio inpainting model are 0.8762 and 0.0255 respectively. This indicates that the Mumford-Shah inpainting model is more effective than Bertalmio inpainting model. The algorithms used in this study can be upgraded to a software framework for commercial use and can be implemented for other kinds of digitized data.

Keywords

• bertalmio model
• image inpainting
• mumford-shah model
• old malay jawi manuscripts