A optical image mosaic methods based on a micro-electro-mechanical system gyroscope sensor mounted with camera rigidly was presented. The relation between the camera rotation angles and the movements of point’s image coordinates is derived based on camera projection model and coordinates transformation. The rotation angle is calculated by the data from gyroscope, thus the search window size could be obtained; then by using the SURF feature extraction algorithm to extract the feature points, the feature matching is conducted in the search window which avoids global searching and accelerates the matching speed by reducing computation. The experiment platform was built to verify the algorithm. The result shows that 31% running time was reduced compared to the traditional global matching.