Engineering　surfaces　always　have　fractal　characteristics　and　now　a　composite　evaluation　method　is　proposed　in　which　the　dual　tree　complex　wavelet　combined　with　fractal　theory　in　this　paper.　By　using　the　dual-tree　complex　wavelets＇　approximate　translation　invariance　and　good　directions,　signal　is　decomposed　into　a　more　delicate　low　and　high　frequency　signals.　Meanwhile,　according　to　the　invariance　of　the　signals＇　fractal　dimension　under　multi-scale,　the　signal＇s　fractal　dimension　is　obtained　by　the　use　of　the　autocorrelation　image　gray　value,　and　the　dual　tree　complex　wavelets＇　decomposition　scale　is　verified　by　calculating　the　different　fractal　dimensions　between　the　high　and　low　frequency　signals　(the　fractal　dimension　distance).　The　simulation　results　show　that　the　datum　can　be　well　extracted　by　the　use　of　the　composite　evaluation　method,　and　the　accuracy　of　decomposition　level　which　is　confirmed　by　the　fractal　dimension　distance　is　verified　by　the　use　of　the　root　mean　square　value.　Two　examples　illustrate　that　reference　and　waviness　of　the　nano-scale　three-dimensional　rough　surfaces　which　have　fractal　characteristic　can　be　well　extracted　by　the　use　of　the　composite　evaluation　method,　which　provides　a　reliable　theoretical　basis　for　the　assessment　of　the　actual　engineering　surface.　©,　2014,　Chinese　Society　for　Measurement.　All　right　reserved.