OCRXBot: Optimizing Image-to-Text Conversion with Tesseract

Text extraction from images is essential for document digitization, automated data entry, and assistive technologies. Traditional OCR systems often struggle with low-quality images and noise, reducing accuracy. To overcome these limitations, a deep learning-based system enhances Tesseract OCR using advanced preprocessing techniques. The method applies grayscale conversion, Gaussian Blur, Otsu’s thresholding, and CLAHE to reduce noise and improve contrast. These preprocessing techniques refine text regions, minimize distortions, and enhance OCR. Instead, a deep CNN was utilized for training the dataset to improve the robustness of text recognition models. Compared to conventional OCR, this approach improves recognition accuracy, adaptability, and efficiency. It bridges the gap between rule-based OCR and intelligent pattern recognition, benefiting document processing, accessibility, and automated data extraction. Future advancements include further dataset optimization, refining preprocessing techniques, and enabling real-time processing for broader applications. KEYWORDS: Tesseract OCR, Deep Learning