With　the　continuous　development　of　machine　learning　in　the　field　of　graphics　and　images,　there　have　been　many　studies　on　the　implementation　methods　of　real　painting　simulation.　Because　there　are　many　places　of　cross　integration　between　different　disciplines,　the　application　scope　of　machine　learning　is　becoming　wider　and　wider.　Therefore,　this　study　uses　the　k-mean　clustering　algorithm　to　segment　and　extract　features　of　painting　works,　and　then　uses　the　existing　graphics　and　image　software　to　simulate　the　nature　of　ink　painting　and　its　diffusion,　and　makes　a　series　of　transformation　synthesis　according　to　the　existing　camera　photos　to　simulate　and　transform　the　color,　line,　texture,　edge　diffusion　and　other　effects　of　the　picture.　Finally,　the　above　results　are　applied　to　the　art　creation　of　traditional　painting,　which　makes　the　original　style　of　traditional　painting　more　realistic.　The　experiment　shows　that　using　machine　learning　method　to　create　traditional　painting　can　realize　the　unity　of　authenticity　and　artistically.　©　2023　IEEE.