Style Transfer of Chinese Opera Character Paintings

Traditional opera character painting blends Chinese color ink painting with conventional opera, requiring both advanced opera art understanding and Chinese painting skills. However, creating these paintings demands specific artistic skills and the ability to integrate elements of opera art. Addressing these challenges is essential for the long-term development of opera character paintings and the promotion of traditional Chinese culture. Recent advancements in artificial intelligence (AI), particularly in image style transfer technology, offer new methods for generating opera character paintings. This paper uses our collected dataset of traditional opera character photographs and their paintings, employing an adversarial generative network (GAN) for implicit feature learning. Consequently, real opera character photographs are transformed into distinctive Chinese color ink paintings. To enhance the quality of the generated images, the encoding process incorporates feature mapping and attention module integration, which help reduce unimportant areas. Additionally, adaptive normalization techniques are used to improve neural network stability, accelerate convergence, and strengthen generalization capabilities. Simulation experiments demonstrate that the proposed method outperforms existing CNN and CycleGAN-based methods, indicates promising transfer performance and highlights the potential for the artistic creation of opera character paintings.