An　improved　algorithm　for　image　mosaic　is　proposed　based　on　traditional　Harris　corner　detection　method.　First,　by　modifying　the　corner　detection　area,　and　corner　response　function,　you　can　detect　the　corners　more　accurately.　Then　use　a　method　based　on　K-d　tree　and　RANSAC　to　match　the　feature　points　and　purify　mismatching　points.　At　last,　using　the　algorithm　based　on　Poisson　equation　to　fuse　the　spliced　images　and　complete　image　mosaic.　The　simulation　shows　that　the　method　can　remove　redundant　feature　points　effectively,　improve　the　image　mosaic　accuracy　and　robustness,　it　is　good　to　complete　the　image　mosaic,　and　has　good　practicability.