Edge　detection　is　widely　used　in　biological　vision　and　computer　vision.　Canny　edge　detection　is　a　common　method　to　locate　the　sharp　intensity　changes　and　seek　object　boundaries.　Nowadays,　there　are　several　improvement　designs　for　Canny　edge　detection　algorithms.　In　this　paper,　the　Gaussian　filtering　is　replaced　by　median　filtering　to　increase　the　noise　robustness　of　Canny　edge　detection.　The　inhabitant　of　the　false　edge　is　achieved　by　implementing　an　improved　Sobel　operator　and　iterative　threshold　filter　method　before　the　non-maximum　suppression　of　the　Canny　operator.　The　final　image　is　obtained　after　threshold　filtering　and　binarization.　The　results　show　that　the　proposed　algorithm　is　more　robust　to　noise　and　can　preserve　more　useful　edge　information　by　suppressing　more　false　edges　than　traditional　Canny　edge　detection.　©　2022　IEEE.