An Adaptive Global Motion Estimation Method Based on Improved SUSAN Algorithm and SIFT Algorithm

In order to improve the real-time of global motion estimation, we proposed a adaptive global motion estimation method based on improved SUSAN algorithm and SIFT algorithm. According to the five matching results before the current matching, the method uses Kalman filter algorithm to predict overlap regions of the current matching two images, and then extracts feature points in the overlapping regions instead of the whole regions of the images. In the part of extracting feature points of the method, improving the SUSAN algorithm according to the geometric characteristics of the SUSAN templates, which improves the speed of extracting feature points. Writing code to implement the method in VS2008, and experimental verification: This method accelerates the executing speed of the algorithm ensuring the accuracy at the same time.