Image feature detection has been developed rapidly in the computer vision, and the SURF feature description can be very stable and fast. RANSAC can estimate the parameters of the model under the condition that the inliers is more than 50%, and it plays a key role in the feature points matching. The surf descriptor of image feature points of detection and description, and then applying the cross matching strategy effectively eliminate wrong matching points, then the use of RANSAC algorithm to estimate the model. Finally, using the linear weighted method for image fusion. This method can be used to match images in low time complexity with the the SURF feature detector and descriptor and RANSAC algorithm.