Skip to Main Content
Engineering surfaces are characterized by the form, waviness and roughness features that are comprised of a range of spatial wavelengths. Filtering techniques are commonly adopted to separate these different wavelength components into well-defined bandwidths for further processing. The Gaussian filtered surface in which a 2D Gaussian filter is employed for surface assessments has been recommended by the ISO 11562-1996 and ASME B46-1995 standards to establish a reference surface. For Gaussian filtering, computational efficiency is a key problem when it is issued on a large set of surface metrology data. In the past this problem was tackled through reducing computation amount by the design and adoption of some fast algorithms. In this paper, a general purpose computing on GPU (GPGPU) framework is discussed to accelerate 2D Gaussian filtering for surface characterization. This framework takes advantage of the GPUpsilas parallel computing ability and has achieved better data efficiency without reducing the computational amount while maintaining the filtering quality. Filtering results and their accuracy from this model have been compared with the results obtained from the MATLAB simulation kits and the satisfied outcomes were observed.