Abstract: With the development of 3D scanning and modeling techniques, 3D models become extremely large. At the same time, intelligent, pervasive computing requires the data to be transmitted as freely and quickly as possible over networks. Traditional progressive transmission methods, whether based on mesh models with arbitrary topological structures or parametric subdivision surfaces, have the problem of huge data transmission and low quality of display effects. To address such problems, a new algorithm is presented. First, we reconstruct the 3D mesh models with displaced butterfly subdivision surfaces, then design a new method to simplify the model dataset, and finally introduce a new progressive transmission algorithm. The algorithm is not only able to decrease the data to be transmitted, but also is effective in improving the display quality.