The aim of this work is to make a comparison of the most current signal processing techniques used to analyze the fringe signal in coherence scanning interferometry (CSI), a major technique for optical surface roughness measurements. We focus here on classical AM-FM signal-processing algorithms such as the Hilbert transform (HT), the five-sample adaptive (FSA), and the continuous wavelet transform (CWT). We have recently also introduced a new family of compact and robust algorithms using the Teager-Kaiser energy operator (TKEO). We propose an improved version of TKEO using a combination of different techniques of pre-filtering and demodulation processing to remove the noise and offset component and to retrieve the fringe envelope to either determine the surface height information or to separate adjacent transparent layers. In particular, as a pre-filtering approach, we have focused on empirical mode decomposition in combination with the Savitzky-Golay filter. An added Gaussian post-filtering is helpful for a precise peak extraction. The experimental results show that TKEO performs better than CWT in terms of computation time and provides a better surface extraction than HT and FSA. Results have been obtained on synthetic and real data taken from a layer of resin on a silicon substrate. (C) 2016 Optical Society of America.
展开▼
