Interference signal restoration for step sample measurement by white light interferometry

White light interferometry (WLI) has currently emerged as one of the mainstream research directions in the optical measurement field, owing to its outstanding performance in nondestructive surface micromorphology measurement. However, when measuring step samples, the complex diffraction effect at the steep step edges gives rise to a bat-wing effect pseudo-impairment in the coherent signal, accompanied by the loss of coherence, reduced signal fidelity, phase ambiguity, and envelope distortion. These issues lead to errors in the height measurement, subsequently resulting in distortions in the reconstruction of the sample morphology. This phenomenon is particularly pronounced when the lateral dimension of the step approaches or reaches the resolution limit (≤ 500 nm in this experiment). In this study, the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise algorithm was adopted to process the designated step edge region, thereby decomposing the distorted original signal into a series of Intrinsic Mode Functions (IMFs). By extracting high-frequency IMFs to filter out low-frequency intensity drift components, the contrast of the coherent signal is enhanced by 91.47%, and the bat-wing effect artifacts are effectively suppressed; the step height measurement relative error is reduced to 3.94%, realizing accurate extraction of the effective interference envelope and high-precision detection of the step structure height.