Brittle Ancient Document Using Adaptive Local Thresholding

The ancient manuscripts must be preserved wholeness and authenticity because there is much important information. The manuscripts are usually kept in libraries and museums for long periods of time. Because of this, the manuscript suffered damages due to a paper that is old. Therefore, prevention needs to be done. One of them by turning the ancient script into a digital image obtained by capturing the image with the camera. However, when an ancient manuscript has become a digital image, the noise on the paper will also be visible. Therefore, image segmentation is needed. Segmentation is the process of separating an object from the background. This research uses five test data and the method used is Mean-C Method, Sauvola Method, and Niblack Method. The evaluation using MSE and PSNR based on ground-truth. Of the three methods, the visual evaluation of the Sauvola Method and the Mean-C Method are good results and the Niblack Method still left many noises. The average measurements of MSE and PSNR from the five data from the Mean-c Method are 2813,71 and 75,58 dB, the results of MSE and PSNR Sauvola Methods are 3308.16 and 70,152 and the last test of the Niblack Method of MSE and PSNR are 8998, 51 and 45.99 dB.