Feature extraction of non-stochastic surfaces using curvelets

Abstract

With the development of precision engineering, higher manufacturing qualities are demanded for advanced optical components. Subsequently characterization of surface topographies is demanded to be more specific and more comprehensive. The methods defined in ISO standards concerns only the overall statistical properties of surfaces, thus they are not applicable to non-stochastic surfaces. The second generation curvelet transform, which provides a sparse representation and good multiscale decomposition performance for curve singularities, is utilized as a powerful tool for the characterization of surface topographies. An effective method is also developed to identify and extract the topography features of interest in the domain of curvelets. Numerical experiments are given to show the effectiveness of this algorithm in sparse representation and feature separation of structured surfaces containing surface waviness, defects, tool marks and irregular scratches.