Ancient Kannada Handwritten Character Recognition from Palm Leaf Manuscripts Using PyTesseract-OCR Technique

Ancient Kannada character recognition is a strenuous task since the Kannada language evolved over various decades, expanding and diversifying the character set. In this research, it is attempted to extract and recognize ancient palm leaf characters using the PyTesseract-OCR engine. The PyTesseract engine is used to denote the optical character recognition (OCR). It can “read” text that has been created into images—a wrapper for the Tesseract optical character recognition (OCR) engine written in Python. Tesseract-OCR uses Google’s text-to-speech voice engine to facilitate optical character recognition. In the present paper, the captured image is further digitized and required to pre-process the ancient inscriptions to recognize the Kannada handwritten characters. The binarization process is used as a processing technique to improve image quality. In the next step, segmentation is carried out with the blob technique, and regional zoning is applied to create the character level. The efficiency of the processed character segmentation strategy as an accuracy of F1-score is 88.3%. Finally, character recognition is carried out, and its precision is above 82.3%. The outcome of this research is being done to contribute to the development of natural language processing research and development in the form of a comprehensive Kannada database.