A Predictive Model for Text Quality Analysis: Case Study

摘要

Text quality is a key aspect of overall print quality. Assessing text quality objectively and quantitatively has remained a challenge, despite our longstanding desire to reach this goal. The range of quality attributes is still seen by many as too broad and the definitions too vague and subjective. In this study, we aim to help overcome these obstacles by exploring whether key attributes exist that can be easily quantified and dependably correlated with subjective perceptions of print quality. If such attributes can be found, we believe a simple predictive model can be developed. For insight into which perceived attributes are critical and to help us select and design objective measurement algorithms, we started by conducting a subjective survey. Guided by the results, we performed quantitative stroke quality measurements and found good correlations between basic stroke properties (e.g., blurriness, stroke width and contrast) and the subjective survey results. We also found that text defects introduced complicating factors into the predictive model. This study provides the foundation for a more comprehensive future study.