Robust estimation of wave-front aberration distribution function using invariant wavelet transform profilometry

The accuracy and stability of the estimation process of phase aberration directly affects the success of machine vision processes. Therefore, accurate modeling and estimation of noise in the process of recovering of aberration distribution function is of prime importance. In this paper, the noise model is considered as an additive Gaussian translation-independent distribution function. To reduce its impacts on estimating the aberration function, non-linear filtering is done in the wavelet space. Thus, the wavelet coefficients are modified in a way that minimizes the effects of the additive Gaussian noise. Then from the modified wavelet coefficients, the phase aberration function is extracted. Experimental results and the empirical validity of this idea are evaluated.

► We present a technique to increase the stability of estimating aberration function.
► The noise is considered as a white noise additive to image brightness function.
► The aberration function is estimated by using the wavelet profilometry technique.
► The noise is reduced by thresholding the wavelet coefficients.
► Compared with Kemao approach, we achieved about 15.8 db improvements in SNR.