A METHOD FOR FAST GAUSSIAN FILTERING （快速高斯滤波的方法）.PDF

A METHOD FOR FAST GAUSSIAN FILTERING FOR 3D ROUGHNESS EVALUATION W. Zeng a, Y. Gao b, *, K. O b , T. Xie a a Huazhong University of Science and Technology, Wuhan, Hubei, 430074, China b Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong * E-mail: meygao@ust.hk Introduction In the assessment of engineering surfaces, it is fundamental to decide correctly a datum line or plane. The precision of datum line or plane directly influences the precision of the assessment of surface parameter. In the assessment of parameters of surface roughness, the method of mean line assessment datum and the one of filtering are often to be used. The mean line assessment datum mainly includes the least square mean line and the arithmetic mean line. The mean line methods have two inherent disadvantages. Firstly, along the overall evaluation length, the assessment datum is not continuous. A step may appear at the crossover region of two adjacent sampling lengths. Naturally, the datum should be a smooth curve. Secondly, the datum greatly relies on the sampling length to extract surface roughness. Distortions in surface roughness assessment may occur if the sampling length is too long or too short. The traditional filtering method mainly relied on 2RC filter. Many instruments of surface roughness in the earlier period use the 2RC filter in the electronic filtering. Due to the inherent nonlinear phase characteristics, the 2RC filter may lead to distortions in a reference datum plane. The 2RC filter has gradually been phased out in recent years. The wavelet method has been widely applied in the areas of signal processing. In surface feature processing, the wavelet datum to be used as an assessmen