Virtual try-on technology has gained increasing attention in recent years, especially in image generation and human modeling. A key challenge in 2D virtual try-on tasks is achieving natural and accurate warping of clothes to fit various human poses. Thin Plate Spline (TPS) warping, widely adopted for its strong geometric warping capabilities, often suffers from distortion and detail loss when handling complex clothing structures. To address these limitations, we propose a warping framework that integrates a feature enhancement mechanism into the TPS-based pipeline. This mechanism enables the model to adaptively focus on body-contact regions such as the shoulders, chest, and waist, enhancing the representation of critical features while mitigating warping artifacts. In addition, we propose a modeling of inter-feature relationships, allowing the network to balance global contour alignment with local detail preservation. Extensive experiments demonstrate that the proposed model achieves superior performance in visual quality, cloth-body alignment, and detail retention, outperforming conventional TPS-based approaches. Our method introduces minimal computational overhead, offering a practical and effective solution for improving warping in virtual try-on systems. The code and models are available at: https://github.com/xinqingting/CST.