A product correlation algorithm of image registration based on feature and spatial-temporal correlation was presented. At first edge features of real-time images and base image are extracted by using Bubble wavelet. Then, the base image is shifted according to position guided by inertia system, in order to make theposition of real-time image in base image to be same. Finally, owing to the direct ratio between probability of registration and area of registration, the sequential product correlation matrix is added, and the registration position is searched in the added matrix. This method has overcome the shortcoming that the probability of registration decreasing with the increasing of the area owing to geometric distortion, and has improved the probability, precision and robustness of registration.