Research on stylization and emotion transfer effect of animated images based on generative adversarial networks

With the rapid development of deep learning technology, generative adversarial networks (GANs) have made significant progress in the field of image generation and stylization. The current cartoon style image generation methods still have limitations, the color is not real, the local details of the image are not in place, etc., to quickly convert the input image into the style output of the anime also need to be combined with deep learning research. The aim of this study is to explore the GAN-based stylization method of animated images and its application effect in emotion transfer. This paper reviews the basic principles and development of generative adversarial networks, and analyzes the advantages and limitations of GAN in image stylization. A GAN-based animated image stylization method is proposed, which combines the emotion theory and realizes the generation of images with different emotion styles by adjusting the network structure and loss function in the generation process. Through comparative experiments, this paper evaluates the stylization effect and the quality of emotion transfer, and verifies the effectiveness of GAN in transferring animation emotion and realizing the conversion of artistic style. GAN can not only efficiently realize the conversion of animation image styles, but also transfer the emotional information in the image to a certain degree, and enhance the visual expression and emotional infectivity. This paper discusses the shortcomings of the existing methods and proposes future research directions, fine control of emotion transfer and multimodal style fusion and so on.