Abstract: The traditional mixture Gaussian model is a commonly used method in background estimation, however, the traditional mixture Gaussian model always allocates a fixed number of Gaussian distribution for each pixel, thus it causes the slowdown of the background germinate speed and the waste of system resource;meanwhile, it exists that the object misjudge phenomenon, namely empty problem, caused by slowdown or demurrage of moving object in the modeling of Gaussian model background.To solve the problem, we propose an effective two-phase method for video images processing.In the first phase, this method automatically adjusts the number of Gaussian distribution based on the priority of each pixel;in the second phase, it firstly partitions the area of each pixel, then uses different update method for object area and non-object area.The experiments show that the two-phase method for video image processing can significantly improve the speed of background model, furthermore efficiently resolve the empty problem in extracting objects, so it verifies the effective of the method.