Benchmarking and Enhancing Bengali OCR: A Hybrid OCR System with Analytic Hierarchy Process-Based Evaluation

This study evaluates Bengali OCR performance using a real-world dataset covering diverse document types. It compares Tesseract and EasyOCR using character- and word-level accuracy measures and examines the effect of image preprocessing. It also proposes an AHP-based hybrid OCR framework that combines outputs from Tesseract, EasyOCR, and Google Vision API to improve recognition performance.