A segmentation free approach for indexing digitized Syriac manuscripts

This paper presents a method to assist the indexation of digitized Syriac manuscripts. Syriac belongs to the Aramaic branch of Semitic languages, it is written from right to left, and the characters hold the feature of being intentionally tilted by an angle of approximately 45°. For the indexation purpose we propose a word spotting approach that should locate all the occurrences of a certain query word image. The method is based on a selective sliding window technique from which directional features are extracted. Matching between features is done using Euclidean distance correspondence. The proposed method does not require any prior information, it is also fully independent of a word to character segmentation algorithm, which would be extremely difficult to realize due to the tilted nature of the handwriting.