Segmentation of Characters in a Handwritten Kannada Document

Objective: The main purpose of the proposed research work is to perform the segmentation of characters from the handwritten Kannada document. The reason behind segmentation is to support the implementation of handwriting recognition system for Kannada language. Methodology: To perform segmentation of characters, input document has to go through gray scale conversion, denoising, contrast normalization and binarization process. Results: Documents collected from ICDAR-2013 and ICDAR-2015 were considered for experiment and obtained 100% accuracy for line segmentation and 96% accuracy for character segmentation. Conclusion: To further improve the efficiency with respect to accuracy of character segmentation, other pre-processing steps like skew detection and correction shall be considered.