Abstract: To solve the problem-great amount of calculation and poor efficiency in present image matching, we proposed a new algorithm combined particle swarm and Topological similarity. In this algorithm we test the characteristic points, compute the contour similarity, and then find out the areas where the image characteristic values distribute obviously. Computing the Topological similarity in the image edge, and optimizing the characteristic values with the particle swarm algorithm, until it iterates successively and produce many-one or many-many relationship. Then transform these complex relationships into one-to-one relationship utilizing the Topological constraint. Compared with the experiment of DPSOHM, the Matching precision greatly improves in this algorithm which resolves the poor efficiency and large error, and greatly reduces the error of the matching points.