The　merits　and　demerits　of　existing　reference　filtering　methods　for　evaluation　of　the　three-dimensional　engineering　surface　roughness　were　discussed.　It　points　out　that　the　Dual-tree　Complex　Wavelet　Transform　(DT-CWT)　is　suitable　for　the　three-dimensional　surface　reference　filter　because　of　its　advantages　of　approximate　translation　invariance　and　improved　directional　selectivity.　However,　its　filtering　is　too　idealistic,　which　will　cause　an　opposition　at　the　point　of　distortion　of　the　filtered　reference.　Therefore,　this　paper　brings　the　airspace　bilateral　filtering　into　the　DT-CWT　transform　to　enhance　its　robustness　at　the　point　of　distortion　of　the　filtered　reference.　The　low-frequency　reference　signal　was　obtained　through　reconstruction　and　each　reconstructed　coefficient　was　bilaterally　filtered,　then,　the　robustness　of　reference　filtering　was　enhanced　by　the　DT-CWT　filtering.　The　simulation　data　show　that　the　robustness　filtering　under　the　conditions　of　abnormal　interference　is　enhanced　by　the　improved　DT-CWT,　the　precision　error　is　0.1832×10-10　m　and　the　reconstruction　error　is　0.0161×10-10　m,　which　is　much　lower　than　that　of　the　traditional　methods.　The　processing　results　for　practical　measurement　data　indicate　that　the　improved　method　well　overcomes　the　edge　effects　and　suppresses　the　effects　of　deep　valley　in　the　surface　on　the　filtering　results.　It　provides　a　new　way　for　evaluation　of　engineering　surface　roughness.