The digitization of architectural heritage for textile applications faces significant challenges, particularly in converting weathered stone relief patterns into machine-readable vector data. This study focuses on extracting high-fidelity geometric features from Shikumen architectural ornamentation, which is characterized by low-contrast edges and high-frequency granular noise. A novel digital workflow is proposed, integrating adaptive image processing algorithms with structural weaving design. First, a Gaussian-weighted bilateral filter suppresses noise while preserving edge discontinuities. Subsequently, a comparative analysis of Sobel, Prewitt, and Canny operators determines that the optimized Canny algorithm (with hysteresis thresholds T_low=0.05, T_high = 0.15) yields the highest feature retention rate. Second, a Grayscale-to-Structure Mapping (GSM) model is developed to translate visual depth into specific Jacquard weave parameters, utilizing an 8-end satin scale to simulate relief effects. The proposed method is validated through the fabrication of a high-density Jacquard fabric (64 ends/cm). Experimental results demonstrate that the extracted vectors achieve a Jaccard Similarity Index of 0.89 against ground truth, and the manufactured fabric exhibits precise structural definition and meets commercial upholstery standards (ASTM), with controlled float lengths (< 4.0 mm). This research provides a robust, quantifiable methodology for the industrial transformation of complex cultural heritage textures into advanced textile products.

digital textile technology, edge detection algorithms, jacquard weaving structure, Shikumen architecture, image segmentation