Seam-aware texture deformation framework for garment design based on thin plate splines

Seamless garment design has gained growing attention in digital garment modeling technologies, as it reduces physical trials and material waste. However, existing approaches heavily rely on high-fidelity 3D garment modeling and periodicity-based texture mapping. To address these challenges, we propose a seam-aware texture deformation framework based on thin plate splines (TPSs), which models texture deformation according to the geometric shapes and spatial relationships of pattern seams. Specifically, a midcurve is constructed to capture the garment sewing shape and guide texture segmentation. Feature pairs between texture and pattern are then established by preserving normalized arc-length ratios and angle-distance relationships. Finally, the TPS method incorporates a regularization term to balance deformation smoothness and boundary geometric fidelity. Experimental results indicate that a regularization coefficient of λ=2.6×{10}⁻² yields effective performance. Across different seam types, cumulative Hu deviations remain below 0.200, Learned Perceptual Image Patch Similarity (LPIPS) values below 0.400, and Hausdorff distances within 20 pixels ( ≈0.339 cm in physical space), demonstrating strong geometric accuracy and perceptual consistency. Remaining variations are primarily attributed to inherent geometric discrepancies between texture segmentation and seam types, revealing the limits of local deformation control. Comparative evaluations further demonstrate that the TPS method outperforms affine and radial basis function approaches in both objective and subjective assessments, with improved shape preservation and seam continuity.