EVALUATION OF LOW-COHERENCE INTERFERENCE FRINGE PARAMETERS BY THE ADAPTIVE WIENER FILTERING METHOD
Low coherence interferometry method is widely used in contactless profilometry and optical coherence tomography (OCT). Low coherence fringes exist within the coherence length of the light source that allows evaluating the distance to the reflecting object under the criterion of maximum fringe visibility at zero optical path difference (OPD) between the object and reference waves. Fringe visibility is represented by the fringe envelope when the incoherent intensity background is removed.
In practice, the background component is significantly variable. In addition, fringe phase discretization step is usually unknown with high accuracy and can be variable, e.g., because of non-uniformity of the OPD change, especially when scanning semitransparent objects with variable properties in depth.
To solve the aforementioned problems, we propose to apply the adaptive Wiener filtering (WF) method. The method is based on adding several fringe signal samples with weighting coefficients and dynamic tuning of the coefficients under criterion of the fringe signal error minimization with respect to a generated reference signal. It has been shown the possibility to extract the estimate of the local fringe phase discretization step from the initial fringe signal itself when this signal is used as the reference one in the adaptive WF. The phase step is calculated via the WF coefficients. When the fringe phase step becomes known, the corresponding reference signal with fixed amplitude is generated. Then the WF coefficients allow calculating fringe envelope with effective suppression of the background component.
The advantage of the proposed WF application consist in absence of initial conditions due to the WF extracts all useful information from the initial signal. Moreover, the WF works with minimal signal samples per fringe period limited by the Nyquist criterion only. This provides high computational speed. Experimental verification of the algorithm has shown the processing speed of real data in time-domain OCT approaching to 1 GVoxels per second when using the conventional computational means.
The work was financially supported by the Russian Science Foundation (Grant No. 19-79-10118).
Igor P. Gurov
File with report