Computational validation study demonstrates high-accuracy segmentation and detection in historical pictographic manuscripts, highlighting end-to-end pipelines for cultural heritage preservation.
Automated analysis of degraded historical pictographic manuscripts is hindered by irregular glyph geometry, non-linear layouts, and fragmented processing pipelines. We present an end-to-end framework that integrates STMS-Net for stroke-aware segmentation, HiST-Glyph for polygon-level detection, a context-aware restoration module, and a spatial-context semantic interpretation network. STMS-Net combines a hierarchical Swin Transformer, dual attention, atrous spatial pyramid pooling, progressive decoding, and high-resolution fusion. HiST-Glyph couples MobileNetV3 and Vision Transformer features with deformable convolution and PSENet localization. Restormer-based reconstruction and graph neural reasoning support structural recovery and contextual interpretation. On Dongba1800, the framework achieved 99.21% accuracy, 98.87% precision, 98.54% sensitivity, 99.45% specificity, a 98.70% F-measure, and a Matthews correlation coefficient of 0.972, outperforming the evaluated baselines. These results support unified analysis of degraded pictographic manuscripts while highlighting the need for broader cross-script validation.
No takes yet. Share an insight, caveat, or question.
Lu et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: