CG-DIQA: No-reference Document Image Quality Assessment Based on Character Gradient

TL;DR

This paper introduces a no-reference document image quality assessment method that leverages character gradient features and OCR accuracy to effectively predict image quality, outperforming existing approaches.

Contribution

The novel approach uses character gradient and MSER-based character patches to estimate document quality without reference images, improving accuracy over prior methods.

Findings

Outperforms state-of-the-art DIQA methods on benchmark datasets

Uses OCR accuracy as a ground-truth metric for quality assessment

Employs character gradient features for effective quality prediction

Abstract

Document image quality assessment (DIQA) is an important and challenging problem in real applications. In order to predict the quality scores of document images, this paper proposes a novel no-reference DIQA method based on character gradient, where the OCR accuracy is used as a ground-truth quality metric. Character gradient is computed on character patches detected with the maximally stable extremal regions (MSER) based method. Character patches are essentially significant to character recognition and therefore suitable for use in estimating document image quality. Experiments on a benchmark dataset show that the proposed method outperforms the state-of-the-art methods in estimating the quality score of document images.