Textile recolorization for nondestructive visual evaluation based on structure-aware curvature-guided image decomposition

Fabric image recolouring offers a non-contact digital strategy for generating alternative colour schemes and evaluating visual design effects without physically altering textile samples. However, fabric images usually contain complex motif structures and dense woven textures, making direct recolouring susceptible to boundary blurring, colour bleeding, and loss of high-frequency texture details. To address these issues, a structure-preserving fabric image recolouring framework is proposed for image-based non-destructive evaluation and design visualisation. The input image is first decomposed into cartoon and texture components using the proposed structure-aware curvature-guided spatial vectorial total variation model, which separates macro-scale motif structures from micro-scale woven details. The cartoon component is then segmented using pattern-adaptive strategies, and selected regions are recoloured in Lab colour space to ensure perceptually consistent colour modification. Finally, the normalised texture component is fused back into the recoloured cartoon image to preserve fabric surface details. Experimental results demonstrate that the proposed method achieves controllable recolouring while maintaining motif integrity and texture fidelity, providing a useful non-destructive tool for fabric colour design and visual evaluation.