Visual background extraction algorithm (ViBe) for foreground detection has the disadvantage that there is ghost and it is difficult to eliminate it for a long time, so an improved visual background extraction algorithm is proposed. Firstly, being introduced in the first sequence with n video frames, the OTSU algorithm can calculate an adaptive threshold for the frame difference method so as to get a more accurate foreground region. Secondly, a sample image is synthesized by using the first n frame images without foreground regions. Finally, the model is initialized with the extended neighborhood range in the synthesized sample images, and the ViBe background model can be updated by the extended domain. The proposed algorithm is compared with various classical algorithms in a large number of video databases. Simulation results show that the improved ViBe algorithm can quickly eliminate the effect of ghost on foreground detection, so the foreground detection is more accurate.