Automated evaluation of historical manuscript text legibility

Establish reliable automated methods for evaluating text legibility in degraded historical manuscript images, supporting task-driven restoration frameworks and evaluation of larger collections.

Background

The paper identifies automated assessment of text legibility as a persistent challenge in document image analysis, particularly for degraded historical manuscripts. Reliable automated evaluation is needed to support restoration systems that optimize for readability and to process larger quantities of cultural heritage data.

The authors note that existing image-quality assessment measures are primarily designed for natural images and may not capture document legibility adequately. They further explain that dedicated evaluation protocols and comprehensive datasets are needed to assess current and future legibility measures, while publicly available datasets supporting such evaluation remain very limited.