In order to analyze engineering surfaces fast and accurately, a new filtering approach based on Butterworth wavelet is proposed. Firstly, the transmission characteristic of a low-pass filter based on a wavelet is discussed,then it is taken as a primary reference to choose an appropriate wavelet base for surface analysis. The construction principle of the Butterworth wavelet filter is briefly introduced, and a fast algorithm for implementing this filter is illuminated. By combining the eminent transmission characteristic and efficient algorithm, the Butterworth wavelet is selected as the analyzing filter to decompose the surface profile with arbitrary orders. Finally,the Butterworth wavelet is also applied to calculating the reference line for profile evaluation and to offering a method to determine decomposed orders. The experimental results indicate that the Butterworth wavelet can achieve the multi-scale analysis of surface profile quickly and accurately and can extract its mean line reliably. The total calculation of this extraction for 11 200 data points only costs 60 ms by general PC and the relative error of Ra obtained by this extracted mean line is only 0.12% larger than the result by Gaussian filter.