Atomic　force　microscope　(AFM),　which　has　nanoscale　precision,　is　a　widely　used　instrument　in　material　science　and　biomedical　science.　Nevertheless,　conventional　AFM　scanning　is　a　time-consuming　procedure.　Compressive　sensing　(CS)　and　undersampling　techniques　have　been　introduced　to　accomplish　fast　AFM　imaging　in　recent　years.　At　present,　existing　undersampled　scan　patterns　cannot　simultaneously　guarantee　imaging　efficiency　and　quality　well.　Therefore,　a　novel　one　called　undersampled　raster　(USR)　scan　is　put　forward　in　this　article.　It　has　higher　imaging　efficiency　than　most　existing　scan　patterns,　which　is　drawn　by　calculating　the　scanning　time　with　the　proposed　estimation　formulas.　Experimental　results　show　that　the　imaging　quality　is　obviously　better　than　using　other　combinations　of　fast　scan　patterns　and　reconstruction　algorithms　when　the　regular　form　of　the　USR　scan　is　employed　with　Total　Variation　Minimization　by　Augmented　Lagrangian　and　Alternating　Direction　Algorithm　(TVAL3).　The　universality　of　this　imaging　scheme　is　verified　by　using　a　variety　of　samples.　In　the　end,　the　applications　of　regular　USR　scan　or　its　variant　for　super-resolution　AFM　imaging　and　predicting　appropriate　sampling　rates　are　proposed.　In　conclusion,　applying　regular　USR　scan　and　TVAL3　algorithm　to　CS-based　AFMcan　effectively　realize　fast　and　high-quality　imaging.