The evaluation method of wear surface topography plays an important role in the evaluation of the wear surface condition and the analysis of the friction characteristics,and the filtering method of the surface topography data is one of the key components of the evaluation method.Four typical robust weight functions,such as Tukey,Hampel,IGGI and QC,were combined with Gaussian filtering to separate the data collected from the surface of adhesive wear and abrasive wear.The robustness and efficiency of the robust Gaussian filter were studied by comparing the reference data of the wear surface obtained by the robust Gaussian filter and the standard Gaussian filter,and analyzing the influence of the high frequency evaluation parameters separated from the wear surface.The robustness of filtering was further discussed in terms of the distribution characteristics of high frequency power spectral density after filtering.The relationship between the outliers and the wear characteristics was discussed.The study shows that Tukey and Hampel robust Gaussian filter have good filtering robustness,and can effectively separate the reference data and highfrequency information containing wear characteristics of the wear surface.The wear characteristics are mainly distributed in the low frequency band of high frequency information.The wear condition and tribological characteristics are closely related to the high frequency information power spectral density function of low frequency band.