Aiming　at　the　problem　of　redundancy　phenomenon　in　the　motor　process　of　image　matching　feature　points,　we　proposed　a　kind　of　global　motion　which　use　Kalman　filter　algorithm　to　estimate　the　overlapping　areas　of　matching　images.　This　algorithm　only　to　extract　the　feature　points　in　the　overlapping　area,　part　of　the　extracted　feature　points　also　proposed　an　effective　combination　of　SUSAN-SURF　algorithm.　SURF　retain　the　high　efficiency　and　SUSAN　has　the　outline　information.　And　then　SURF　algorithm　is　effectively　improved　by　using　KNN　to　speed　up　image　matching,　determine　the　matching　point　to　realize　image　registration.　Finally　the　algorithm　was　verified　by　experiment,　this　global　motion　method　can　under　the　premise　in　accuracy,　improve　the　real-time　performance.