A Two-Step Robust Filter for Mean Line Extraction Based on the Median Filter

A robust Gaussian regression filter and a robust spline filter have been proposed in ISO/TS 16610 to reduce the influence of the deep valleys and high peaks in profiles. However, the differences of the transmission characteristics between the robust Gaussian regression filter and robust spline filter will lead to different evaluation results because of the different calculation methods. It´s notable that median process is utilized in both of the robust filters. It shows that the median process is the key point for obtaining robust characteristics. In this paper, a novel two step robust filter is proposed based on the median filter. The calculation process is separated into two steps including the robustness step and the smoothing step. The robustness step is to obtain the robust characteristics of the profile and the smooth step is to eliminate the short-wavelength component of the profile. The experiment showed that an appropriate window length of the median filter will make it possible to not only obtain the robust characteristics but also preserve the long-wavelength components. Spline filter is recommended for the smoothing step because of its advantage of little edge effect. Its effectiveness and application were verified by the experiments.