Painting style rendering is an important technology of non-photorealistic rendering in computer graphics. The style transition from one image to another can be seen as a material transition. The purpose of texture migration is to preserve the Semantic information of the original image by extracting and constraining the texture of the original image. With the increasing popularity of painting art, the style rendering of painting art has become a research hotspot. However, in order to render the artistic style of painting into the desired style, users usually need to master and adjust a large number of parameters, which is very inconvenient for their use. Deep neural network has a strong ability of autonomous learning and data processing. Its birth promotes the rapid development of natural language processing, image generation, semantic segmentation and other fields. Therefore, the research on image stylization based on deep learning has achieved good results. This paper proposes a depth neural network algorithm, which can not only generate effective painting pictures, but also render different styles of target objects in the original image.