“Digital city” in a broad sense refers to urban informatization. That is to say, the digital city will gradually realize the informatization of the urban national economy and society by building infrastructure platforms such as broadband multimedia information network and geographic information system, integrating urban information resources, establishing e-government, e-commerce, labor, social insurance, and other information communities, to make the city in an invincible position in the competition of the information age. The virtual reality technology of urban three-dimensional landscape, or the establishment of a virtual city, is an important part of the vast project of digital city construction. This research mainly discusses the rendering and optimization algorithm of digital city 3D artistic landscape based on virtual reality. The article uses image stitching technology. Clean up unnecessary furniture in the CAD floor plan, and try to keep only the wall structure. Export drawings and save as dwg files for 3DMAX modeling. For adjacent overlapping images, the common features of the overlapping parts are determined by image matching, so that the images taken at different shooting positions and different viewing directions with a small angle of intersection are calculated to be unified. Take the photo coordinate system, and then find out the overlapping range of each direction for splicing. In the actual data acquisition process, most of the spatial data such as existing maps are used. It mainly includes topographic maps and the latest planning drawings, as well as planning and design drawings of various residential areas. In addition to determining the specific location of each feature, various maps are used to make a surface image (texture) of an urban area based on these data using Photoshop. The editing of roaming space is the key to realizing 3D panoramic virtual city browsing. Combined with the navigation of 3D panoramic map, using key technologies such as hotspot connection and command button, the 3D panoramic images collected in different places in the city can be effectively integrated into one, assisting with various multimedia elements, such as background music, animated videos, and language commentary, show the true appearance of the virtual city in all directions. Compared with the SURF algorithm, the SIFT algorithm extracts more feature points and more correctly matched point pairs, and the minimum time is 0.524 s. This research will contribute to the construction of the three-dimensional landscape of the digital city.