A Unified Framework for Degraded Thai Historical Document Image Restoration

Binarization method is the key process to restore degraded historical document image. In this paper, the framework for degraded Thai historical document image restoration is proposed. The proposed framework consists of three stage including image filtering stage, local-based thresholding stage, and cluster analysis stage. Image filtering stage aims to eliminate some noises by using Wiener filter. Local-based thresholding stage aims to calculate the optimal threshold of a local block by using Niblack’s methods. Cluster analysis stage aims to improve the quality of binary image by using Kim’s method. The experiments are implemented by using Matlab and conducted on real degraded Thai historical document image dataset which is provided by Nation Library of Thailand. The experimental results are evaluated by using three widely used indices including precision, recall and f-index. The experimental results show that the proposed framework outperforms four classical binarization methods.