抄録
Fast M-Estimation Gaussian Filter (FMGF) is the robust filter with the low computational cost of which transmission characteristics agrees with Gaussian filter's one if and only if without an outlier. 1D-FMGF is verified to possess all of the three major characteristics of filters for linear roughness measurement (“characteristic agree with Gaussian filter”, “high-speed” and “robustness”) is verified. In this paper, FMGF is applied to the areal surface texture. When 1D-FMGF is applied repeatedly in perpendicular directions, FMGF meets the requirement “directional characteristic”. However, if with an outlier, FMGF does not meet the requirement of “directional characteristic”. Then, this paper proposes new FMGF by changing some processes. New FMGF is confirmed that it could obtain “directional characteristic” if outlier exists, too. And finally it was confirmed that the FMGF could obtain all of the four major characteristics of filters for areal surface texture.
