Abstract: Image art style transfer can apply the style of one image to the content of another image, thus creating an image with a new style, which is widely used in the fields of art creation, cartoon animation production, de-sign and advertising, providing artists, designers and advertisers with a new way of creation and expres-sion. This paper reviews the literature at home and abroad, and summarizes the research status of image art style transfer technology. On this basis, focusing on the field of deep learning, the principles, advantages and disadvantages of convolutional neural network and Transformer, generative adversarial network, and diffusion model methods are analyzed respectively. Then, the performance of several mainstream migration algorithms is verified and analyzed using Microsoft COCO and WikiArt as test datasets. In addition, the main problems in the current research such as image generation quality and fidelity, controllability of style generalization, computational efficiency, and the future development directions such as mixed control of diverse artistic styles, multi-modal collaboration, human-machine collaboration, and real data open sce-narios are also discussed, which provides a reference for the further research of image stylization technol-ogy.