Image　registration　is　of　prime　importance　for　image　mosaic.　It　is　the　process　of　geometrically　aligning　one　image　to　another　image　of　the　same　scene　taken　from　different　time,　from　different　viewpoints　or　by　different　sensors.　In　general,　image　registration　needs　a　lot　of　time.　This　paper　reduces　the　computation　time　from　two　aspects.　For　coarse　matching,　the　paper　combines　normalized　cross　correlation　(NCC)　matching　and　sequential　similarity　detection　algorithm　(SSDA)　to　reduce　unnecessary　computation.　For　precise　matching,　by　introducing　an　outlier　constraint,　the　paper　terminates　computation　early　by　comparing　outliers.　By　adopting　these　improvements　there　may　be　a　saving　of　computation　time　of　about　92.57%　and　8.17%　respectively.　©2010　IEEE.