Image blending based on style transfer
Hailong Zhu, Xiangchu Feng, Jiao Zhe · 2025
This paper proposes a style transfer-based image blending method that effectively combines style transfer and Poisson editing by introducing gradient loss, enabling cross-style image blending. Compared to traditional image blending methods, the proposed approach significantly improves blending quality when handling images with large style differences, producing images with consistent style and rich details. Experimental results demonstrate that, in tasks involving image blending with significant style differences, the proposed method outperforms existing techniques and better preserves the style features and detailed structures of the images.