Improving Text Recognition in Natural Scenes using Optimized OCR Techniques

This paper presents a method for optimizing Optical Character Recognition (OCR) performance for the Telugu script in natural scenes. The method focuses on minimizing the Character Error Rate (CER) through the fine-tuning of Tesseract OCR settings, the implementation of advanced preprocessing techniques, and the incorporation of a dictionary-based post-processing correction mechanism. The proposed approach was evaluated against existing methods, demonstrating superior accuracy in text recognition for the Telugu script. The results emphasize the method's potential for application to other underrepresented languages with complex scripts, paving the way for more effective and inclusive OCR technologies. The research provides a foundation for further advancements in OCR systems, particularly in addressing the unique challenges associated with non-Latin scripts in diverse environments.