Segmentation of Handwritten Sanskrit Words Using Image-Processing Techniques

The development of efficient character recognition frameworks for handwritten Sanskrit words is inherently dependent on the underlying segmentation techniques employed. This dependency can be attributed to the intricacies in the Sanskrit script—including the similarity between characters, the variation in writing styles, and the presence of a voluminous number of characters, modifiers, and conjuncts. Furthermore, hybrid segmentation techniques need to be used for line-level segmentation and word-level segmentation. Therefore, further research can be accelerated with the introduction of novel techniques for the segmentation of Sanskrit words. The proposed methodology outlines a systematic technique for Sanskrit word segmentation, employing methods including thresholding, zone-based classification, and projection profiles, among others, and also accounts for the numerous types of characters and modifiers involved, and accordingly includes relevant sections that extensively discuss the segmentation of modifiers. The authors seek to make this implementation publicly available to promote further collaborative research in this domain.