Image　registration　is　an　important　technology　of　image　processing　and　has　a　broad　application.　Good　adaptability,　computing　speed　and　registration　accuracy　are　the　basic　requirements　in　image　registration.　In　recent　years,　the　image　registration　algorithm　based　on　local　invariant　features　is　becoming　a　hot　spot　of　research,　such　as　the　SIFT　method　is　one　of　them.　But　the　traditional　SIFT　algorithm　has　computational　complexity　and　the　high　real-time　performance.　SURF　algorithm　as　a　new　feature　extraction　method,　possess　a　small　amount　of　calculation　and　quick　speed,　but　the　large　amount　of　feature　points　are　create　false　match　easily.　To　solve　these　problems,　this　paper　proposed　an　effective　combination　of　SUSAN-SURF　algorithm,　which　retain　the　high　efficiency　of　SURF　and　the　outline　information　of　SUSAN.　And　then　SURF　algorithm　is　effectively　improved　by　using　KNN　to　speed　up　image　matching.　Finally　the　algorithm　was　verified　by　experiment,　SUSAN-SURF　can　under　the　premise　in　accuracy,　improve　the　real-time　performance.　©　2014　IEEE.