Boosting Optical Character Recognition: A Super-Resolution Approach

TL;DR

This paper presents a super-resolution framework that significantly improves OCR accuracy on low-resolution text images, achieving results close to high-resolution images, and was successful in the ICDAR2015 competition.

Contribution

The paper introduces a novel super-resolution approach that enhances OCR performance on low-resolution images, demonstrating competitive results in a major international competition.

Findings

OCR accuracy improved to 77.19% from low-resolution images

Super-resolution performance is comparable to high-resolution images

Achieved winning entry in ICDAR2015 Text Image Super-Resolution challenge

Abstract

Text image super-resolution is a challenging yet open research problem in the computer vision community. In particular, low-resolution images hamper the performance of typical optical character recognition (OCR) systems. In this article, we summarize our entry to the ICDAR2015 Competition on Text Image Super-Resolution. Experiments are based on the provided ICDAR2015 TextSR dataset and the released Tesseract-OCR 3.02 system. We report that our winning entry of text image super-resolution framework has largely improved the OCR performance with low-resolution images used as input, reaching an OCR accuracy score of 77.19%, which is comparable with that of using the original high-resolution images 78.80%.