Automated stratification and noise adaptive parameter tuning for ancient stone inscription text extraction

Ancient stone inscriptions constitute a significant archive of cultural heritage, yet their interpretation is hindered by surface degradation, stone texture, illumination variability, and inscribed depth. A review of prior art highlights the need for a unified pre-processing pipeline that addresses variability in imaging features by adapting parameters. In this study, a two-phase framework is proposed. In Phase I, inscription images are stratified using Elbow-based K-Means clustering on discriminative image features, grouping images with similar features. In Phase II, the Dynamically Parameterized Image Pre-processing Pipeline dynamically tunes fundamental pre-processing parameters based on noise estimation. This noise-adaptive mechanism allows optimal enhancement regardless of illumination, texture, or surface degradation. Quantitative evaluation metrics demonstrate that the proposed method outperforms static pre-processing approaches in contrast enhancement, background reduction, and text readability. The proposed framework provides a scalable pre-processing basis for reliable digital and downstream inscription analysis.