Abstract: Image segmentation is an important direction in image understanding, computer vision and character recognition. Thresholding is a simple, efficient and widely used method in image segmentation. In addition, it helps to reduce the complexity of data and it simplifies recognition and classification. It discriminates a background from the object pixels depending on the suitable threshold value selected. Restoration of broken characters in digital image of historical documents is important factor to survive it from losing historical information. Otsu is one of the best thresholding techniques because of its robustness and speed in partitioning background from the object. This is done by increasing insularity factor between the classes. In this study, we attempted to improve image restoration of broken image characters and subsequently made a comparison among some automatic thresholding techniques such as Kittler, Maxentropy and Otsu. In this process, we employed Hausdorff Distance (HD) as a measure of performance evaluation. Our experimental results showed that Otsu thresholding technique outperforms others.