A new style transfer method based on color prioritization
Weiyu Peng, Xuezhong Qian, Wei Song · 2023
Style transfer has been widely used and developed in cultural entertainment industry. However, the traditional neural style transfer cannot perfectly transfer the color style of artistic images. To address this critical issue, we propose a color-first realtime style transfer framework. On the one hand, we ensure adequate transmission of color by giving priority to color transfer; On the other hand, we design a color perceptual loss to reduce the loss of color style. Qualitative and quantitative experiments demonstrate that compared with other state-of-the-art methods, our framework can effectively transfer the color style of artistic images with higher accuracy and richness, as well as good visual perception.