Atomic　Force　Microscopy　(AFM)　is　one　of　the　most　popular　and　advanced　tools　for　ultra　high-resolution　imaging　and　nanomanipulation　of　nano-scale　matter.　But　AFM　imaging　typically　takes　a　long　time.　High-speed　and　high　precision　AFM　measurement　has　attracted　wide　attention　in　recent　several　years.　In　traditional　AFM,　simple　reduction　in　the　number　of　measurement　points　may　lose　essential　sample　topography　information.　To　resolve　such　problems,　an　AFM　image　reconstruction　method　based　on　Compressed　Sensing　(CS)　theory　is　applied　to　reduce　image　acquisition　time　without　cutting　down　the　image　quality.　The　benefit　of　using　CS　approach　in　AFM　is　shortening　the　imaging　time,　minimizing　the　interaction　with　the　sample,　and　finally　avoiding　sample　damage　in　AFM.　Three　kinds　of　testing　samples　with　high　and　low　frequency　components　were　examined　by　a　scanning　electron　microscope　(SEM)　and　by　AFM.　An　orthogonal　Matching　Pursuit　(OMP)　algorithm　is　employed　to　reconstruct　an　AFM　image　with　different　sampling　rates.　Subsequently　the　reconstruction　results　of　sample　topography　images　are　analyzed　and　evaluated.　Using　the　CS　approach　in　AFM　can　greatly　improve　the　AFM　imaging　process.　Experimental　results　show　that　the　obtained　reconstructed　images　have　different　resolution　and　quality,　depending　on　the　surface　morphology　of　the　sample　and　sampling　rates.