Abstract: Contour matching is an integral part in image processing. Analyzing the limitation of the conventional contour matching methods which take a long time to search, an improved algorithm is proposed for contour matching. In the algorithm the contours of the template image and of the image to be matched are extracted separately so as to calculate the curvature of every point on contours, and the candidate point is selected from those points on contour for satisfying the threshold. Then the eigenvectors are built at the candidate point and several points on its both sides. The comparability measurement function is built according to the Euclidean distance, and the global optimization problem is solved by differential evolution algorithm. Comparative experiment results showed that the method proposed has quick optimizing speed and high matching ability.