In　this　paper,　an　image　denoising　model　which　embeds　intrinsic　mode　function　into　Perona-Malik　model　is　proposed.　Firstly,　the　image　is　decomposed　into　Intrinsic　Mode　Functions　(IMFs)　by　using　empirical　mode　decomposition　technique;　each　of　IMFs　captures　the　feature　information　under　different　scales.　Secondly,　the　first　and　second　IMFs　are　embedded　into　Perona-Malik　model.　Experimental　results　indicate　that　this　method　is　more　efficient　than　Perona-Malik　model　in　removing　Gaussian　noise.　Moreover,　this　method　can　remove　salt　and　pepper　noise.