Intuitive Hairstyle Transfer Editing via User-Guided Mask and Hairstyle Patch Integration

We propose a novel interactive hair editing system that seamlessly integrates user manipulation into a StyleGAN-based hairstyle transfer framework. Our approach centers on incorporating intuitive user editing intentions into both the structure and appearance blending stages of conventional hairstyle transfer systems. The structure blending stage integrates user-modified hair masks into latent space optimization, facilitating natural hair region adjustments, while the appearance blending stage in- corporates hair patches generated from user sketches to achieve fine-grained control over color and texture expressions. Experimental results on the FFHQ dataset, complemented by comprehensive user studies, demonstrate that our method significantly outperforms state-of-the-art approaches across mul- tiple metrics, including naturalness, user intention reflection, satisfaction, and both structural and color accuracy. Notably, our interactive editing paradigm substantially enhances the expressiveness and flex- ibility of hair editing by enabling users to realize desired hairstyles with unprecedented precision and intuition.