Handwritten Image OCR for Tamil Languages Using Tesseract OCR

Project Synopsis & Overview

This project develops a Handwritten Image Optical Character Recognition (OCR) system for recognizing and converting handwritten Tamil text from images into editable digital text using Tesseract OCR. The system collects a diverse dataset of handwritten Tamil text and performs preprocessing techniques such as noise removal, binarization, skew correction, resizing, and normalization to improve image quality. Character segmentation is performed to separate text into lines, words, and individual characters. Relevant features are extracted using geometric, structural, gradient, and Histogram of Oriented Gradients (HOG) features. The Tesseract OCR engine is trained using the prepared handwritten Tamil dataset to learn different character patterns and handwriting styles. During recognition, the trained model identifies Tamil characters, groups them into words, and generates editable digital text. Post-processing techniques such as spell checking and language-based correction can be used to reduce recognition errors. The system also supports evaluation using Character Error Rate (CER), Word Error Rate (WER), and sentence-level accuracy. The proposed system aims to provide an efficient solution for digitizing handwritten Tamil documents for archival, academic, personal, and administrative applications.

Complete Technical Specifications