Recognition of Ancient Stone Inscription Characters using Normalized Positional Distance Metric Features

Recognition of inscribed characters on stone is an essential part of our work to reveal information's of our ancestors. The concept of Optical Character Recognition is not adopted as the rock surface becomes texture of an image posing challenges in recovering the text. In this paper, we proposed a new feature called Positional Distance Metric which was independent to any language script to address the various issues occurred on stone inscribed images. Then Normalized Positional Distance Metric (NPDM) feature is computed and combined with structural and regional features to yield a better recognition rate. We repeated this procedure for all the training images and finally populated in to the list. Nearest Neighbor classifier is used for subsequent classification and recognition of characters. Experiments are performed on 350 characters and showed significant recognition rate.