OSA's Digital Library

Optics Express

Optics Express

  • Editor: Andrew M. Weiner
  • Vol. 22, Iss. 4 — Feb. 24, 2014
  • pp: 4738–4750
« Show journal navigation

Generalized phase-shifting algorithm for inhomogeneous phase shift and spatio-temporal fringe visibility variation

Rigoberto Juarez-Salazar, Carlos Robledo-Sanchez, Fermin Guerrero-Sanchez, and A. Rangel-Huerta  »View Author Affiliations


Optics Express, Vol. 22, Issue 4, pp. 4738-4750 (2014)
http://dx.doi.org/10.1364/OE.22.004738


View Full Text Article

Acrobat PDF (6670 KB)





Browse Journals / Lookup Meetings

Browse by Journal and Year


   


Lookup Conference Papers

Close Browse Journals / Lookup Meetings

Article Tools

Share
Citations

Abstract

A cascade least-squares scheme for wrapped phase extraction using two or more phase-shifted fringe-patterns with unknown and inhomogeneous surface phase shift is proposed. This algorithm is based on the parameter estimation approach to process fringe-patterns where, except for the interest phase distribution that is a function of the space only, all other parameters are functions of both space and time. Computer simulations and experimental results show that phase computing is possible even when an inhomogeneous phase shift is induced by nonlinearity of the piezoelectric materials or miscalibrated phase shifters. The algorithm’s features and its operating conditions will been discussed. Due to the useful properties of this algorithm such as the robustness, computational efficiency, and user-free execution, this proposal could be used in automatic applications.

© 2014 Optical Society of America

1. Introduction

Phase-shifting techniques are an important toolbox for extracting the wrapped phase from phase-shifted fringe-patterns [1

1. D. Malacara, M. Servin, and Z. Malacara, Interferogram Analysis for Optical Testing, 2nd ed. (Taylor and Francis, 2005). [CrossRef]

3

3. X. Xu, G. Lu, G. Han, F. Gao, Z. Jiao, and D. Li, “Phase stitching and error correction in aperture synthesis for generalized phase-shifting interferometry,” Appl. Opt. 52, 4864–4870 (2013). [CrossRef] [PubMed]

]. The wrapped phase extraction is a key step to phase demodulation processing in numerous non-contact measurement systems such as fringe-projection, optical interferometry, synthetic aperture radar, and other similar applications. In such measurement systems, the physical parameter to be measured (strain, temperature, refraction index, topography, etc.) is encoded as a phase distribution in a cosinusoidal way [4

4. G. Rajshekhar and P. Rastogi, “Fringe analysis: Premise and perspectives,” Opt. Lasers Eng. 50, iii–x (2012). [CrossRef]

], namely
Ik(p)=ak(p)+bk(p)cosΦk(p),withΦk(p)=ϕ(p)+δk(p),
(1)
for k = 0, 1,··· ,K − 1, where K is the number of fringe-patterns, Ik(p) is a spatial intensity distribution, p = (x, y) is a two-dimensional spatial variable, ak(p) is the background light, bk(p) is the modulation light, Φk(p) is the encoded phase, ϕ(p) is the phase function of interest to be recovered, and δk(p) is the phase shift function.

The standard phase-shifting approach relies mainly on three assumptions [5

5. J. H. Bruning, D. R. Herriott, J. E. Gallagher, D. P. Rosenfeld, A. D. White, and D. J. Brangaccio, “Digital wavefront measuring interferometer for testing optical surfaces and lenses,” Appl. Opt. 13, 2693–2703 (1974). [CrossRef] [PubMed]

]. First, the background and modulation functions are equal in all fringe-patterns; i.e., they are not dependent on k. Second, the phase shift function is commonly homogeneous in space, i.e., the phase shift does not depend on p (the phase shift function is a set of non-tilted planes). And third, the phase shift function must be known a priori.

The K-step algorithms [5

5. J. H. Bruning, D. R. Herriott, J. E. Gallagher, D. P. Rosenfeld, A. D. White, and D. J. Brangaccio, “Digital wavefront measuring interferometer for testing optical surfaces and lenses,” Appl. Opt. 13, 2693–2703 (1974). [CrossRef] [PubMed]

] and the least-squares method [6

6. C. J. Morgan, “Least-squares estimation in phase-measurement interferometry,” Opt. Lett. 7, 368–370 (1982). [CrossRef] [PubMed]

, 7

7. J. E. Greivenkamp, “Generalized data reduction for heterodyne interferometry,” Opt. Eng. 23350–352 (1984). [CrossRef]

] are associated with the standard approach. Computationally, this approach is the most efficient. Unfortunately, to know the phase shift with high accuracy, expensive shifter devices are required and calibration procedures are necessary [8

8. Y.-Y. Cheng and J. C. Wyant, “Phase shifter calibration in phase-shifting interferometry,” Appl. Opt. 24, 3049–3052 (1985). [CrossRef] [PubMed]

10

10. W. Li and X. Su, “Real-time calibration algorithm for phase shifting in phase-measuring profilometry,” Opt. Eng. 40, 761–766 (2001). [CrossRef]

]. To avoid these restrictions, the generalized phase-shifting approach was suggested [11

11. G. Lai and T. Yatagai, “Generalized phase-shifting interferometry,” J. Opt. Soc. Am. A 8, 822–827 (1991). [CrossRef]

].

The generalized phase-shifting approach does not explicitly require knowing the phase shift function. This approach includes a stage dedicated to the estimation of the phase shift. The Carrè algorithms may be the first phase-shifting techniques in this category [12

12. P. Carre, “Installation et utilisation du comparateur photohlectrique et interfhrentiel du bureau international des poids et mesures,” Metrologia 2, 13–23 (1966). [CrossRef]

, 13

13. P. Hariharan, B. F. Oreb, and T. Eiju, “Digital phase-shifting interferometry: a simple error-compensating phase calculation algorithm,” Appl. Opt. 26, 2504–2506 (1987). [CrossRef] [PubMed]

]. However, they are restricted to the case when the phase shift function is a linear function of k. In other words, the phase shift must be non-tilted planes separated one from another by a constant distance as shown in Fig. 1(a).

Fig. 1 Phase shift functions in the generalized phase-shifting approach. Homogeneous phase shift: (a) linear, and (b) nonlinear on k. Inhomogeneous phase shift: (c) nonlinear on k but linear on p, and (d) nonlinear on both k and p.

Another case is when the phase shift is homogeneous but is a nonlinear function of k as shown in Fig. 1(b). The most generalized phase-shifting algorithms fall into this group [11

11. G. Lai and T. Yatagai, “Generalized phase-shifting interferometry,” J. Opt. Soc. Am. A 8, 822–827 (1991). [CrossRef]

, 14

14. C. T. Farrell and M. A. Player, “Phase-step insensitive algorithms for phase-shifting interferometry,” Meas. Sci. Technol. 5(6), 648 (1994). [CrossRef]

19

19. R. Juarez-Salazar, C. Robledo-Sanchez, C. Meneses-Fabian, F. Guerrero-Sanchez, and L. A. Aguilar, “Generalized phase-shifting interferometry by parameter estimation with the least squares method,” Opt. Lasers Eng. 51(5), 626–632 (2013). [CrossRef]

].

In some optical setups, it is difficult to keep a homogeneous phase shift. For example, the translation of the reference mirror of an interferometer cannot be exactly perpendicular to the optical axis [20

20. M. Chen, H. Guo, and C. Wei, “Algorithm immune to tilt phase-shifting error for phase-shifting interferometers,” Appl. Opt. 39, 3894–3898 (2000). [CrossRef]

22

22. K. Patorski, A. Styk, L. Bruno, and P. Szwaykowski, “Tilt-shift error detection in phase-shifting interferometry,” Opt. Express 14, 5232–5249 (2006). [CrossRef] [PubMed]

]. For this case, the phase shift function consists of a set of tilted planes as shown in Fig. 1(c). Interesting schemes to solve this problem have been developed [23

23. F. Zeng, Q. Tan, H. Gu, and G. Jin, “Phase extraction from interferograms with unknown tilt phase shifts based on a regularized optical flow method,” Opt. Express 21, 17234–17248 (2013). [CrossRef] [PubMed]

26

26. O. Soloviev and G. Vdovin, “Phase extraction from three and more interferograms registered with different unknown wavefront tilts,” Opt. Express 13, 3743–3753 (2005). [CrossRef] [PubMed]

].

Due to the development of advanced technologies (e.g. liquid crystal spatial light modulators) and new experimental designs [27

27. C. Meneses-Fabian and U. Rivera-Ortega, “Phase-shifting interferometry by wave amplitude modulation,” Opt. Lett. 36, 2417–2419 (2011). [CrossRef] [PubMed]

,28

28. C. Robledo-Sanchez, R. Juarez-Salazar, C. Meneses-Fabian, F. Guerrero-Sánchez, L. M. A. Aguilar, G. Rodriguez-Zurita, and V. Ixba-Santos, “Phase-shifting interferometry based on the lateral displacement of the light source,” Opt. Express 21, 17228–17233 (2013). [CrossRef] [PubMed]

], the requirements of phase-shifting algorithms have increased. Particularly, it is necessary to process phase-shifted fringe-patterns with unknown inhomogeneous phase shifts that depend on both p and k [29

29. K. Hibino, B. F. Oreb, D. I. Farrant, and K. G. Larkin, “Phase-shifting algorithms for nonlinear and spatially nonuniform phase shifts,” J. Opt. Soc. Am. A 14, 918–930 (1997). [CrossRef]

] as shown in Fig. 1(d). Additionally, other desirable properties for phase-shifting techniques are: a reduced number of necessary fringe-patterns, and the capacity to handle spatio-temporal visibility (i.e. the background and modulation lights are functions of both p and k) [30

30. M. Afifi, K. Nassim, and S. Rachafi, “Five-frame phase-shifting algorithm insensitive to diode laser power variation,” Opt. Commun. 197, 37–42 (2001). [CrossRef]

].

At this point, an apparent problem arises. If an arbitrary shift (changing spatially and temporally in an unknown manner) is assumed, it introduces an extra fringe-pattern. The extra fringe-pattern can cause confusion regarding what part of the encoded phase belongs to the interest phase distribution and what part belongs to the phase shift. Moreover, conceptually, the phase shift is a necessary “reference phase” useful for wrapped phase extraction. Therefore, as an arbitrary phase shift is no longer a “reference”, the phase extraction in the mentioned conditions could be out of the phase-shifting framework.

In order to address this situation, the phase-shifting concept is stated as follows. In phase-shifting, the phase distribution of interest ϕ(p) is equal in all fringe-patterns, and the phase shift δk(p) varies from one to another. In other words, ϕ(p) and δk(p) are, respectively, the static and dynamic parts of the encoded phase as can be seen in Fig. 2.

Fig. 2 From a set of phase-shifted fringe-patterns modelled by Eq. (1), the interest phase distribution ϕ(p) and the phase shift δk(p) can be determined, respectively, as the dynamic and static parts of the encoded phase Φk(p) = ϕ(p) + δk(p).

Indeed, for the continuous case, by applying the derivative of the encoded phase with respect to k, the interest phase distribution is suppressed and the remainder is the phase step αk(p), i.e.,
ddkΦk(p)=ddkδk(p)=αk(p).
(2)
Thus, the wrapped phase extraction is possible if αk(p) ≠ 0. This is a basic assumption in the phase-shifting approach. Then, the phase shift is reconstructed by
δk(p)=δ0+0kα(p)d.
(3)
Finally, the interest phase distribution ϕ(p) is recovered by using the phase shift δk(p) computed by Eq. (3). So the functions ϕ(p) and δk(p) are clearly distinguished from one another (even when δk(p) is nonlinear unknown function of both p and k) and the phase computing is possible in the context of phase-shifting. For the discrete case, Eqs. (2) and (3) are the respective finite differences and cumulative sum.

In this work, we present an extension of the previously reported generalized phase-shifting algorithm [19

19. R. Juarez-Salazar, C. Robledo-Sanchez, C. Meneses-Fabian, F. Guerrero-Sanchez, and L. A. Aguilar, “Generalized phase-shifting interferometry by parameter estimation with the least squares method,” Opt. Lasers Eng. 51(5), 626–632 (2013). [CrossRef]

]. This scheme can handle spatio-temporal visibility (i.e. the background and modulation lights are surfaces changing for each interferogram) but it assumes homogeneous nonlinear phase shift of k, Fig. 1(b). For this study, an advanced phase shift estimation stage is developed. Thus, the proposed algorithm can efficiently handle inhomogeneous nonlinear phase shifts of both p and k, Fig. 1(d). The beneficial properties of the algorithm (such as only two or more fringe-pattern requirements, robustness, user-free execution, and computational efficiency) are kept. The results from computer simulation and experimental tests validate this proposal.

2. Theoretical principles

Hereafter in this paper, the following notation is used. For simplicity, all functions of p are defined in a discretized rectangular domain. Thus, they are considered M × N matrices and the variable p is not written down for brevity. For a M × N matrix 𝔸 with MN and rank(𝔸) = N, the notation 𝔸 = (𝔸T 𝔸)−1𝔸T denotes the least-squares inverse of the matrix 𝔸.

The wrapped phase recovered by the proposed generalized phase-shifting algorithm is carried out through three stages: fringe-pattern normalization, phase shift estimation, and wrapped phase extraction.

2.1. Fringe-pattern normalization

The normalization procedure consists of removing the background ak and modulation bk lights from the fringe-patterns Ik described by Eq. (1). For this, we used the parameter estimation approach [19

19. R. Juarez-Salazar, C. Robledo-Sanchez, C. Meneses-Fabian, F. Guerrero-Sanchez, and L. A. Aguilar, “Generalized phase-shifting interferometry by parameter estimation with the least squares method,” Opt. Lasers Eng. 51(5), 626–632 (2013). [CrossRef]

]. Specifically,
a˜k=𝔸a𝔸aIk,
(4a)
b˜k2=2𝔸b𝔸b(Ika˜k)2,
(4b)
where ãk and k are the approximations of the parameters ak and bk, respectively. The columns of the matrices 𝔸a and 𝔸b are the basis functions employed.

Depending on the applications, the basis functions can be truncated Taylor polynomials, Fourier series, piecewise polynomials (splines) of appropriate degree, B-splines, etc. For the simulation and experiments in this work, we used a 2nd-degree polynomial to estimate ãk (a 2nd-order approximation of a Gaussian profile) and, a 4th-degree polynomial to estimate b˜k2.

Finally, the normalized fringe-patterns are computed by
I¯k=sat(Ika˜kb˜k)cos[ϕ+δk],
(5)
where the saturation function sat(·) bounds the argument to the interval [−1, 1]. For more details on this normalization procedure, the reader is referred to Ref. [19

19. R. Juarez-Salazar, C. Robledo-Sanchez, C. Meneses-Fabian, F. Guerrero-Sanchez, and L. A. Aguilar, “Generalized phase-shifting interferometry by parameter estimation with the least squares method,” Opt. Lasers Eng. 51(5), 626–632 (2013). [CrossRef]

].

2.2. Phase shift estimation

We consider the addition Ak and subtraction Sk of adjacent normalized fringe-patterns defined, respectively, by
Ak=I¯k1+I¯k=2cosαk2cos[ϕ+δk1+δk2],
(6a)
Sk=I¯k1I¯k=2sinαk2sin[ϕ+δk1+δk2],
(6b)
for k = 1, 2,··· ,K − 1, where αk = δkδk−1 (a discrete version of Eq. (2) for Δk = 1) is the phase step between the fringe-patterns Ik and Ik−1. From the above equations, by applying some trigonometric identities and a few algebraic operations, we can reach:
Ak21=cosαk+2cos2αk2ηk,
(7a)
1Sk2=cosαk+2sin2αk2ηk,
(7b)
where ηk = cos(2ϕ + δk−1 + δk). It is possible to estimate the term cosαk from any above equation by considering that the factor ηk is noise with zero mean and amplitude of 2cos2(αk/2) or 2sin2(αk/2), respectively. However, in some situations, one is more convenient than another as shown in Fig. 3(a). Particularly, for αk ∈ [0, π/2], it is appropriate to choose 1Sk2, Eq. (7b), because the noise amplitude is low (and it is higher in Ak21), and vice versa for αk ∈ [π/2, π].

Fig. 3 (a) The data Ak21 and 1Sk2, given by Eqs. (7)(a) and (7)(b), for the estimation of the term cosαk. For αk ∈ [0, π/2] it is convenient to choose 1Sk2 because the noise amplitude is lower, and vice versa for αk ∈ [π/2, π]. (b) Switch functions by hyperbolic tangent approximation of unit step functions. The constant ω sets the smoothness of the transition at ck = 0. In this plot, ω = 15. (c) Equivalent data to polynomial fitting by the procedure described with Eq. (9).

Since it is not possible to choose a priori the data Ak21 or 1Sk2, we carried out a polynomial fitting for these two data. Mathematically,
cAk=𝔸c𝔸c(Ak21),
(8a)
cSk=𝔸c𝔸c(1Sk2),
(8b)
where cAk and cSk are the fitted polynomials which approximate the term cosαk from Ak21 and 1Sk2, respectively. The columns of the matrix 𝔸c are the basis functions of the polynomial used. Particularly, in this work, a 4th-degree Taylor polynomial is used.

Now, to select between the polynomials cAk and cSk, we consider the approximation Γ(·) = [1 + tanh(·)]/2 of the unit step function, Fig. 3(b), as
cosαk=cSkΓ(ωck)+cAkΓ(ωck)=ck+12(cSkcAk)tanh(ωck),
(9)
where ck = (cAk + cSk)/2, and ω is an appropriate constant (experimentally, we chose ω = 15). Thus, a smooth transition between cSk and cAk at ck = 0 is obtained.

It is worth mentioning that, instead of carrying out a polynomial fitting procedure to any data set as shown in Fig. 3(a), the procedure described by Eq. (9) is similar to performing the polynomial fitting to the equivalent data shown in Fig. 3(c).

2.3. Wrapped phase extraction

Finally, the wrapped phase distribution is computed for each point p as
ϕw=arctan(ζ/ξ),
(11)
where the quantities ζ = sinϕ and ξ = cosϕ are obtained by the least-squares method [6

6. C. J. Morgan, “Least-squares estimation in phase-measurement interferometry,” Opt. Lett. 7, 368–370 (1982). [CrossRef] [PubMed]

, 7

7. J. E. Greivenkamp, “Generalized data reduction for heterodyne interferometry,” Opt. Eng. 23350–352 (1984). [CrossRef]

] as
[ξζ]=𝔸ϕ[I0I1IK1]T,
(12)
with
𝔸ϕ=[cosδ0sinδ0cosδK1sinδK1].
(13)
It is worth mentioning that the above matrix 𝔸ϕ is, in general, different for each point p. Therefore, this last stage is time-consuming because a matrix inversion must be performed for each point to solve Eq. (12). However, the runtime can be reduced by parallel computing because the wrapped phase extraction is pointwise. Particularly, the wrapped phase extraction is faster when the phase shift is homogeneous (non-tilted planes). In this case, the matrix 𝔸ϕ is the same for all points p and a single matrix inversion is necessary.

This proposed generalized phase-shifting algorithm to inhomogeneous phase shift and spatio-temporal visibility is depicted by the block diagram shown in Fig. 4.

Fig. 4 Proposed inhomogeneous generalized phase-shifting algorithm by a cascade least-squares scheme. (a) Normalization stage. (b) Phase shift estimation stage. (c) Wrapped phase extraction stage.

3. Computer simulation and experimental results

In this section, the functionality of the proposal is verified by a computer simulation. Then, the feasibility and robustness are tested by processing experimental phase-shifted fringe-patterns.

3.1. Computer simulation

We consider the phase step α1 and the phase distribution ϕ given, respectively, by
α1=π[12+13(x2y2)],
(14)
ϕ=6peaks(500)+12(x2+y2)+1,
(15)
for x, y ∈ [−1, 1]. The command peaks(500) is defined in MATLAB software and returns a linear combination of Gaussian functions. The plots of Eqs. (14) and (15) are shown in the first column of Fig. 5.

Fig. 5 Computer simulation. (1st column) Phase step α1, phase distribution ϕ to be recovered and its wrapped version. (2nd column) Background a0 and modulation b0 lights for the first fringe-pattern I0. Similarly, the 3rd column shows a1, b1 and I1. (4th column) Normalized fringe-patterns, computed phase step, and recovered wrapped phase.

Now, two noisy fringe-patterns are simulated as
Ik=ak+bkcos(ϕ+δk)+ρk,
(16)
for k = 0, 1, where ρk is a Gaussian noise with zero mean and standard deviation σ (10%). The respective background and modulation lights are given by
a0=15(x2+y2+1),b0=x3+y2+2,a1=20(y2x2+1)+5x,b1=y3x2+2.5.
(17)
The functions ak, bk, and Ik are shown in the second and third columns of Fig. 5 for k = 1, 2, respectively.

The two simulated fringe-patterns Ik, Figs. 5(j) and 5(k), are processed by the proposed algorithm. The resulting normalized fringe-patterns Īk, phase step α1, and the recovered wrapped phase are shown in the last column of Fig. 5.

By direct comparison between Figs. 5(i) and 5(l), we can see that a correct solution was reached. This claim is verified by the error information shown in Fig. 6. The first column of Fig. 6 shows the absolute error from estimations of phase step, Fig. 6(a), and the wrapped phase, Fig. 6(c). The second column of Fig. 6 shows the respective error histograms (normalized in frequency).

Fig. 6 Computer simulation. (1st column) Absolute error from the phase step and wrapped phase estimations. (2nd column) The respective error histograms.

3.2. Experimental results

The feasibility of the proposed algorithm was examined by processing experimental fringe-patterns recorded from the Twyman-Green Interferometer shown in Fig. 7(a). A collimated laser beam was obtained by a He–Ne laser source (wavelength λ = 633 nm), a spatial filter (objective microscope and pinhole), and a collimating lens (focal length f = 0.5 m).

Fig. 7 (a) Optical setup for the phase-shifting experiments. Pictographic description of the piezoelectric control signals to generate phase shifts: (a) homogeneous, (b) inhomogeneous tilted planes, and (c) inhomogeneous surfaces.

In the first interferometer’s arm, the translation of the flat mirror M1 is driven by two piezo-electric devices, PZT1 and PZT2. Thus, if the control signals for these devices are equal, Fig. 7(b), a homogeneous phase shift is induced. Otherwise, Fig. 7(c), an inhomogeneous tilted planes phase shift is generated and the tilting is controlled by the difference between the control signals. Moreover, if the control signals are impulsive, Fig. 7(d), an inhomogeneous surface phase shift is induced due to this mechanical perturbation.

A distorted wavefront was generated by inserting a phase object in the second interferometer’s arm. The fringe-patterns observed on the plane OP were recorded by using a gray-scale camera sensor (1824 × 1418 pixels with 8-bit pixel depth). Notice that the phase object is not previously characterized. However, it is measured by first phase-shifting with a conventional homogeneous phase shift. Thus, when inhomogeneous phase shifts are considered, the results must be consistent with the first measure.

The proposed algorithm requires only two or more fringe-patterns to work. However, we considered three fringe-patterns in order to test the phase shift estimation quality later. Accordingly, three phase-shifting experiments were carried out. The induced phase shifts are, respectively: homogeneous (i.e., non-tilted planes), Fig. 8 (1st row), inhomogeneous tilted planes, Fig. 8 (2nd row), and inhomogeneous surfaces, Fig. 8 (3rd row).

Fig. 8 Experimental results. Wrapped phase computed from three phase-shifted fringe-patterns when the phase shift is: (1st row) homogeneous (non-tilted planes), (2nd row) inhomogeneous tilted planes, and (3rd row) inhomogeneous surfaces. (1–3rd columns) Fringe-patterns to be processed. (4th column) Computed phase shift. (5th column) Recovered wrapped phase. (6th column) Resulting background light.

There are several evaluation methods to test the performance of generalized phase-shifting algorithms [31

31. J. Li, Y. Wang, X. Meng, X. Yang, and Q. Wang, “An evaluation method for phase shift extraction algorithms in generalized phase-shifting interferometry,” J. Opt. 15(10), 105408 (2013). [CrossRef]

]. In this work, the proposed algorithm is verified by using the background-phase correlation [32

32. H. Guo, “Blind self-calibrating algorithm for phase-shifting interferometry by use of cross-bispectrum,” Opt. Express 19, 7807–7815 (2011). [CrossRef] [PubMed]

]. The principle of background-phase correlation is as follows. If the phase shift δ is obtained with an error ε ≪ 1 (i.e., we have δ̃ = δ + ε), the least-squares method to phase extraction solves the problem (at least three fringe-patterns are required):
I=a+bcos(ϕ+δ˜)a˜+bcos(ϕ+δ),
where the computed background ã = aεbsin(ϕ + δ) is correlated with the phase function ϕ by the term εbsin(ϕ + δ). Thus, a simple way to test the phase shift estimation quality is by computing the correlation coefficient between the reconstructed background light and the fringe-patterns.

The resulting background light for each phase-shifting experiment is shown in Fig. 8 (6th column). In each experiment, the reconstructed background light is compared with the fringe-patterns and the three respective correlation coefficients are computed. The maximum correlation values are 0.5618 (first experiment), 0.5579 (second experiment), and 0.6271 (third experiment). The accuracy levels reached in all experiments are similar (the computed correlation coefficient is 0.5925 ± 0.0346).

4. Discussion

4.1. Reference frame in phase-shifting

It seems that the extracted wrapped phase maps (5th column in Fig. 8) do not describe the same object phase because these maps are not equal. Since the measured phase object is the same in all measurements, the recovered wrapped phases must match in some sense. Indeed, as phase-shifting techniques measure only the relative phase distribution, it can be transformed by translations and/or rotations without affecting the topography of the recovered phase distribution. That is rather an effect due to the established reference frame. In particular, the “reference point” in phase-shifting is given by the phase shift δk(p) at k = 0 as is shown below.

The wrapped data of the 5th column in Fig. 8 are unwrapped by using the phase-unwrapping method based on network flow programming [33

33. M. Costantini, “A novel phase unwrapping method based on network programming,” IEEE Trans. Geosci. Remote Sens. 36(3), 813–821 (1998). [CrossRef]

]. The respective results (which will be labeled as ϕ1, ϕ2, and ϕ3) are shown in Figs. 9(a)–9(c). The differences ϕ2ϕ1 and ϕ3ϕ1 are shown in Figs. 9(d) and 9(g), respectively. We can see that these differences are planes. This claim is verified by fitting linear functions, Figs. 9(e) and 9(h), where the fitting error, Figs. 9(f) and 9(i), present average values of 4.5 × 10−15 and −3.3 × 10−14, respectively. Although the difference ϕ3ϕ2 is not shown, it can be proved that such difference is also a plane. Accordingly, the experimental wrapped phase maps in Fig. 8 actually describe the same topography.

Fig. 9 (1st row) Unwrapped phases ϕ1, ϕ2, and ϕ3, respectively, of the wrapped phase maps shown in the 5th column of Fig. 8. (2nd row) The difference ϕ2ϕ1, its fitted plane, and the fitting error (4.5 × 10−15 average value). (3rd row) The difference ϕ3ϕ1, its fitted plane, and the fitting error (−3.3 × 10−14 average value).

Notice that, if the initial phase is known (e.g. the last induced phase shift), it can be included as δ0(p) in Eq. (10). This subtracts this cumulated phase from the final wrapped phase map. In any case, the proposed phase-shifting algorithm can extract the interest wrapped phase distribution up to an initial phase shift δ0(p) as occurs with any phase-shifting method.

4.2. Spatio-temporal visibility fluctuations

The spatio-temporal visibility is addressed by the normalization stage in the algorithm’s scheme. This good feature was shown by computer simulation. But, this property could not be confirmed in the experimental evaluations because the employed optical setup does not exhibit such visibility fluctuations. However, we believe that similar results can be achieved in experimental data frames.

4.3. Algorithm performance

The proposed algorithm is based on the computationally efficient least-squares method. The suggested scheme shown in Fig. 4 provides high robustness and easy implementation. Among its processing stages, the wrapped phase extraction is the most computationally intensive because it requires a matrix inversion for each pixel of the fringe-pattern. However, this pointwise computation also allows parallel computing. By using a 2.5 GHz laptop, the computer time required to obtain the results shown in Figs. 5 and 8 is shown in Fig. 10.

Fig. 10 Computing time required for processing: (a) Two 500 × 500 fringe-patterns with inhomogeneous nonlinear phase shift (Fig. 5). (b)–(d) Three 1824 × 1418 fringe-patterns with: homogeneous (constant), inhomogeneous linear, and inhomogeneous nonlinear phase shifts, respectively (Fig. 8). The hardware used was a 2.5 GHz laptop.

From the first bar of the plot in Fig. 10, we can see that the phase extraction from two fringe-patterns of moderate size is carried out in less that one second. For bigger fringe-patterns, quasi-dynamic measurements are still possible as is the case of an optical workshop (e.g., interferometrical monitoring from 1824 × 1418 pixels fringe-patterns every 5 seconds). Notice that this algorithm is very fast when the particular homogeneous phase shift is involved (Fig. 10(b)) because a single matrix inversion in the wrapped phase extraction stage is required.

On the other hand, the algorithm settings (mainly basis functions and order of the polynomials) are established before the algorithm is running. Further user intervention is not required. Consequently, due to the low hardware requirements and user-free execution of the proposal, it could be used in automatic applications.

The simplicity of the suggested algorithm allows its implementation into dedicated hardware such as DSP (digital signal processor). This will reduce the computing time notably. Thus, this algorithm could be used in real-time applications.

4.4. Operating requirements

The proposed algorithm is highly flexible for processing phase-shifted fringe-patterns. However, to obtain the optimal algorithm performance, there are three main issues to keep in mind.

First, the fitting procedure by using the least-squares method is efficient when the data frames are corrupted with additive symmetrically distributed zero mean noise. This method is applied to obtain the functions ak, bk and αk where the additive terms cos(ϕ + δk) and cos(2ϕ + δk−1 + δk) (with their respective amplitudes) are seen as noise. Thus, many fringes (open and closed in any combination) across the recorded intensities Ik are required in order to satisfy both symmetric distribution and zero mean conditions. The property of additivity of the noise does not allow processing speckled fringe-patterns because speckle noise is multiplicative. However, the method could be applicable if pre-filtering of the fringe-patterns is accomplished.

Second, the estimation of the functions ak, bk and αk requires appropriate basis functions. Fortunately, there are a variety of options from which we can choose for each particular application. For example, the basis functions can be polynomials (such as truncated Taylor and Fourier series, Seidel, Zernike, etc.), splines (piecewise polynomials) of appropriate degree, B-splines, etc. This flexibility allows the proposed algorithm to be implemented in many applications.

And third, the phase steps αk are computed as the argument of a cosine function, Eq (8). Therefore, the phase steps are recovered without ambiguity if αk lies within the interval (0, π) rad. Nevertheless, the phase shift δk, the cumulative sum of the phase steps, may be greater than 2π as long as the matrix 𝔸ϕ of Eq. (13) satisfies rank(𝔸ϕ) = 2. If the phase steps exceed the interval (0, π), they are wrapped. We believe that this issue can be overcome by an additional unwrapping procedure; however, this possibility is left as future work.

5. Conclusion

A generalized inhomogeneous phase-shifting algorithm for wrapped phase extraction working with two or more fringe-patterns was proposed. The algorithm is based on a cascade least-squares estimation scheme. This proposal can handle fringe-patterns with spatio-temporal visibility and inhomogeneous surface phase shift. The algorithm features and its operating conditions have been discussed.

Computer simulations and experimental results have shown the feasibility of this algorithm. Moreover, the obtained results have shown that the wrapped phase computing is possible even when an inhomogeneous phase shift is induced (e.g., by nonlinearity of piezoelectric materials or miscalibrated phase shifters) and large visibility changes have occurred. The computational efficiency and user-free execution of this algorithm allows it to be implemented in automatic applications.

Acknowledgments

R. Juarez-Salazar is grateful for the scholarship from the Consejo Nacional de Ciencia y Tecnología (CONACYT), México. C. Robledo-Sanchez thanks the support from CONACYT. The authors thank N. Keranen for her advice on wording. A special gratitude to the anonymous reviewers for their useful recommendations to improve this work.

References and links

1.

D. Malacara, M. Servin, and Z. Malacara, Interferogram Analysis for Optical Testing, 2nd ed. (Taylor and Francis, 2005). [CrossRef]

2.

A. Patil and P. Rastogi, “Moving ahead with phase,” Opt. Lasers Eng. 45, 253–257 (2007). [CrossRef]

3.

X. Xu, G. Lu, G. Han, F. Gao, Z. Jiao, and D. Li, “Phase stitching and error correction in aperture synthesis for generalized phase-shifting interferometry,” Appl. Opt. 52, 4864–4870 (2013). [CrossRef] [PubMed]

4.

G. Rajshekhar and P. Rastogi, “Fringe analysis: Premise and perspectives,” Opt. Lasers Eng. 50, iii–x (2012). [CrossRef]

5.

J. H. Bruning, D. R. Herriott, J. E. Gallagher, D. P. Rosenfeld, A. D. White, and D. J. Brangaccio, “Digital wavefront measuring interferometer for testing optical surfaces and lenses,” Appl. Opt. 13, 2693–2703 (1974). [CrossRef] [PubMed]

6.

C. J. Morgan, “Least-squares estimation in phase-measurement interferometry,” Opt. Lett. 7, 368–370 (1982). [CrossRef] [PubMed]

7.

J. E. Greivenkamp, “Generalized data reduction for heterodyne interferometry,” Opt. Eng. 23350–352 (1984). [CrossRef]

8.

Y.-Y. Cheng and J. C. Wyant, “Phase shifter calibration in phase-shifting interferometry,” Appl. Opt. 24, 3049–3052 (1985). [CrossRef] [PubMed]

9.

X. Chen, M. Gramaglia, and J. A. Yeazell, “Phase-shift calibration algorithm for phase-shifting interferometry,” J. Opt. Soc. Am. A 17, 2061–2066 (2000). [CrossRef]

10.

W. Li and X. Su, “Real-time calibration algorithm for phase shifting in phase-measuring profilometry,” Opt. Eng. 40, 761–766 (2001). [CrossRef]

11.

G. Lai and T. Yatagai, “Generalized phase-shifting interferometry,” J. Opt. Soc. Am. A 8, 822–827 (1991). [CrossRef]

12.

P. Carre, “Installation et utilisation du comparateur photohlectrique et interfhrentiel du bureau international des poids et mesures,” Metrologia 2, 13–23 (1966). [CrossRef]

13.

P. Hariharan, B. F. Oreb, and T. Eiju, “Digital phase-shifting interferometry: a simple error-compensating phase calculation algorithm,” Appl. Opt. 26, 2504–2506 (1987). [CrossRef] [PubMed]

14.

C. T. Farrell and M. A. Player, “Phase-step insensitive algorithms for phase-shifting interferometry,” Meas. Sci. Technol. 5(6), 648 (1994). [CrossRef]

15.

X. Chen, M. Gramaglia, and J. A. Yeazell, “Phase-shifting interferometry with uncalibrated phase shifts,” Appl. Opt. 39, 585–591 (2000). [CrossRef]

16.

K. Larkin, “A self-calibrating phase-shifting algorithm based on the natural demodulation of two-dimensional fringe patterns,” Opt. Express 9, 236–253 (2001). [CrossRef] [PubMed]

17.

A. Patil and P. Rastogi, “Approaches in generalized phase shifting interferometry,” Opt. Lasers Eng. 43, 475–490 (2005). [CrossRef]

18.

Z. Wang and B. Han, “Advanced iterative algorithm for phase extraction of randomly phase-shifted interferograms,” Opt. Lett. 29, 1671–1673 (2004). [CrossRef] [PubMed]

19.

R. Juarez-Salazar, C. Robledo-Sanchez, C. Meneses-Fabian, F. Guerrero-Sanchez, and L. A. Aguilar, “Generalized phase-shifting interferometry by parameter estimation with the least squares method,” Opt. Lasers Eng. 51(5), 626–632 (2013). [CrossRef]

20.

M. Chen, H. Guo, and C. Wei, “Algorithm immune to tilt phase-shifting error for phase-shifting interferometers,” Appl. Opt. 39, 3894–3898 (2000). [CrossRef]

21.

A. Dobroiu, D. Apostol, V. Nascov, and V. Damian, “Tilt-compensating algorithm for phase-shift interferometry,” Appl. Opt. 41, 2435–2439 (2002). [CrossRef] [PubMed]

22.

K. Patorski, A. Styk, L. Bruno, and P. Szwaykowski, “Tilt-shift error detection in phase-shifting interferometry,” Opt. Express 14, 5232–5249 (2006). [CrossRef] [PubMed]

23.

F. Zeng, Q. Tan, H. Gu, and G. Jin, “Phase extraction from interferograms with unknown tilt phase shifts based on a regularized optical flow method,” Opt. Express 21, 17234–17248 (2013). [CrossRef] [PubMed]

24.

J. Xu, Q. Xu, and L. Chai, “Tilt-shift determination and compensation in phase-shifting interferometry,” J. Opt. A Pure Appl. Opt. 10(7), 075011 (2008). [CrossRef]

25.

J. Xu, Q. Xu, and L. Chai, “Iterative algorithm for phase extraction from interferograms with random and spatially nonuniform phase shifts,” Appl. Opt. 47, 480–485 (2008). [CrossRef] [PubMed]

26.

O. Soloviev and G. Vdovin, “Phase extraction from three and more interferograms registered with different unknown wavefront tilts,” Opt. Express 13, 3743–3753 (2005). [CrossRef] [PubMed]

27.

C. Meneses-Fabian and U. Rivera-Ortega, “Phase-shifting interferometry by wave amplitude modulation,” Opt. Lett. 36, 2417–2419 (2011). [CrossRef] [PubMed]

28.

C. Robledo-Sanchez, R. Juarez-Salazar, C. Meneses-Fabian, F. Guerrero-Sánchez, L. M. A. Aguilar, G. Rodriguez-Zurita, and V. Ixba-Santos, “Phase-shifting interferometry based on the lateral displacement of the light source,” Opt. Express 21, 17228–17233 (2013). [CrossRef] [PubMed]

29.

K. Hibino, B. F. Oreb, D. I. Farrant, and K. G. Larkin, “Phase-shifting algorithms for nonlinear and spatially nonuniform phase shifts,” J. Opt. Soc. Am. A 14, 918–930 (1997). [CrossRef]

30.

M. Afifi, K. Nassim, and S. Rachafi, “Five-frame phase-shifting algorithm insensitive to diode laser power variation,” Opt. Commun. 197, 37–42 (2001). [CrossRef]

31.

J. Li, Y. Wang, X. Meng, X. Yang, and Q. Wang, “An evaluation method for phase shift extraction algorithms in generalized phase-shifting interferometry,” J. Opt. 15(10), 105408 (2013). [CrossRef]

32.

H. Guo, “Blind self-calibrating algorithm for phase-shifting interferometry by use of cross-bispectrum,” Opt. Express 19, 7807–7815 (2011). [CrossRef] [PubMed]

33.

M. Costantini, “A novel phase unwrapping method based on network programming,” IEEE Trans. Geosci. Remote Sens. 36(3), 813–821 (1998). [CrossRef]

OCIS Codes
(050.5080) Diffraction and gratings : Phase shift
(120.2650) Instrumentation, measurement, and metrology : Fringe analysis
(120.3180) Instrumentation, measurement, and metrology : Interferometry
(120.5050) Instrumentation, measurement, and metrology : Phase measurement

ToC Category:
Instrumentation, Measurement, and Metrology

History
Original Manuscript: November 7, 2013
Revised Manuscript: December 19, 2013
Manuscript Accepted: January 7, 2014
Published: February 21, 2014

Citation
Rigoberto Juarez-Salazar, Carlos Robledo-Sanchez, Fermin Guerrero-Sanchez, and A. Rangel-Huerta, "Generalized phase-shifting algorithm for inhomogeneous phase shift and spatio-temporal fringe visibility variation," Opt. Express 22, 4738-4750 (2014)
http://www.opticsinfobase.org/oe/abstract.cfm?URI=oe-22-4-4738


Sort:  Author  |  Year  |  Journal  |  Reset  

References

  1. D. Malacara, M. Servin, Z. Malacara, Interferogram Analysis for Optical Testing, 2nd ed. (Taylor and Francis, 2005). [CrossRef]
  2. A. Patil, P. Rastogi, “Moving ahead with phase,” Opt. Lasers Eng. 45, 253–257 (2007). [CrossRef]
  3. X. Xu, G. Lu, G. Han, F. Gao, Z. Jiao, D. Li, “Phase stitching and error correction in aperture synthesis for generalized phase-shifting interferometry,” Appl. Opt. 52, 4864–4870 (2013). [CrossRef] [PubMed]
  4. G. Rajshekhar, P. Rastogi, “Fringe analysis: Premise and perspectives,” Opt. Lasers Eng. 50, iii–x (2012). [CrossRef]
  5. J. H. Bruning, D. R. Herriott, J. E. Gallagher, D. P. Rosenfeld, A. D. White, D. J. Brangaccio, “Digital wavefront measuring interferometer for testing optical surfaces and lenses,” Appl. Opt. 13, 2693–2703 (1974). [CrossRef] [PubMed]
  6. C. J. Morgan, “Least-squares estimation in phase-measurement interferometry,” Opt. Lett. 7, 368–370 (1982). [CrossRef] [PubMed]
  7. J. E. Greivenkamp, “Generalized data reduction for heterodyne interferometry,” Opt. Eng. 23350–352 (1984). [CrossRef]
  8. Y.-Y. Cheng, J. C. Wyant, “Phase shifter calibration in phase-shifting interferometry,” Appl. Opt. 24, 3049–3052 (1985). [CrossRef] [PubMed]
  9. X. Chen, M. Gramaglia, J. A. Yeazell, “Phase-shift calibration algorithm for phase-shifting interferometry,” J. Opt. Soc. Am. A 17, 2061–2066 (2000). [CrossRef]
  10. W. Li, X. Su, “Real-time calibration algorithm for phase shifting in phase-measuring profilometry,” Opt. Eng. 40, 761–766 (2001). [CrossRef]
  11. G. Lai, T. Yatagai, “Generalized phase-shifting interferometry,” J. Opt. Soc. Am. A 8, 822–827 (1991). [CrossRef]
  12. P. Carre, “Installation et utilisation du comparateur photohlectrique et interfhrentiel du bureau international des poids et mesures,” Metrologia 2, 13–23 (1966). [CrossRef]
  13. P. Hariharan, B. F. Oreb, T. Eiju, “Digital phase-shifting interferometry: a simple error-compensating phase calculation algorithm,” Appl. Opt. 26, 2504–2506 (1987). [CrossRef] [PubMed]
  14. C. T. Farrell, M. A. Player, “Phase-step insensitive algorithms for phase-shifting interferometry,” Meas. Sci. Technol. 5(6), 648 (1994). [CrossRef]
  15. X. Chen, M. Gramaglia, J. A. Yeazell, “Phase-shifting interferometry with uncalibrated phase shifts,” Appl. Opt. 39, 585–591 (2000). [CrossRef]
  16. K. Larkin, “A self-calibrating phase-shifting algorithm based on the natural demodulation of two-dimensional fringe patterns,” Opt. Express 9, 236–253 (2001). [CrossRef] [PubMed]
  17. A. Patil, P. Rastogi, “Approaches in generalized phase shifting interferometry,” Opt. Lasers Eng. 43, 475–490 (2005). [CrossRef]
  18. Z. Wang, B. Han, “Advanced iterative algorithm for phase extraction of randomly phase-shifted interferograms,” Opt. Lett. 29, 1671–1673 (2004). [CrossRef] [PubMed]
  19. R. Juarez-Salazar, C. Robledo-Sanchez, C. Meneses-Fabian, F. Guerrero-Sanchez, L. A. Aguilar, “Generalized phase-shifting interferometry by parameter estimation with the least squares method,” Opt. Lasers Eng. 51(5), 626–632 (2013). [CrossRef]
  20. M. Chen, H. Guo, C. Wei, “Algorithm immune to tilt phase-shifting error for phase-shifting interferometers,” Appl. Opt. 39, 3894–3898 (2000). [CrossRef]
  21. A. Dobroiu, D. Apostol, V. Nascov, V. Damian, “Tilt-compensating algorithm for phase-shift interferometry,” Appl. Opt. 41, 2435–2439 (2002). [CrossRef] [PubMed]
  22. K. Patorski, A. Styk, L. Bruno, P. Szwaykowski, “Tilt-shift error detection in phase-shifting interferometry,” Opt. Express 14, 5232–5249 (2006). [CrossRef] [PubMed]
  23. F. Zeng, Q. Tan, H. Gu, G. Jin, “Phase extraction from interferograms with unknown tilt phase shifts based on a regularized optical flow method,” Opt. Express 21, 17234–17248 (2013). [CrossRef] [PubMed]
  24. J. Xu, Q. Xu, L. Chai, “Tilt-shift determination and compensation in phase-shifting interferometry,” J. Opt. A Pure Appl. Opt. 10(7), 075011 (2008). [CrossRef]
  25. J. Xu, Q. Xu, L. Chai, “Iterative algorithm for phase extraction from interferograms with random and spatially nonuniform phase shifts,” Appl. Opt. 47, 480–485 (2008). [CrossRef] [PubMed]
  26. O. Soloviev, G. Vdovin, “Phase extraction from three and more interferograms registered with different unknown wavefront tilts,” Opt. Express 13, 3743–3753 (2005). [CrossRef] [PubMed]
  27. C. Meneses-Fabian, U. Rivera-Ortega, “Phase-shifting interferometry by wave amplitude modulation,” Opt. Lett. 36, 2417–2419 (2011). [CrossRef] [PubMed]
  28. C. Robledo-Sanchez, R. Juarez-Salazar, C. Meneses-Fabian, F. Guerrero-Sánchez, L. M. A. Aguilar, G. Rodriguez-Zurita, V. Ixba-Santos, “Phase-shifting interferometry based on the lateral displacement of the light source,” Opt. Express 21, 17228–17233 (2013). [CrossRef] [PubMed]
  29. K. Hibino, B. F. Oreb, D. I. Farrant, K. G. Larkin, “Phase-shifting algorithms for nonlinear and spatially nonuniform phase shifts,” J. Opt. Soc. Am. A 14, 918–930 (1997). [CrossRef]
  30. M. Afifi, K. Nassim, S. Rachafi, “Five-frame phase-shifting algorithm insensitive to diode laser power variation,” Opt. Commun. 197, 37–42 (2001). [CrossRef]
  31. J. Li, Y. Wang, X. Meng, X. Yang, Q. Wang, “An evaluation method for phase shift extraction algorithms in generalized phase-shifting interferometry,” J. Opt. 15(10), 105408 (2013). [CrossRef]
  32. H. Guo, “Blind self-calibrating algorithm for phase-shifting interferometry by use of cross-bispectrum,” Opt. Express 19, 7807–7815 (2011). [CrossRef] [PubMed]
  33. M. Costantini, “A novel phase unwrapping method based on network programming,” IEEE Trans. Geosci. Remote Sens. 36(3), 813–821 (1998). [CrossRef]

Cited By

Alert me when this paper is cited

OSA is able to provide readers links to articles that cite this paper by participating in CrossRef's Cited-By Linking service. CrossRef includes content from more than 3000 publishers and societies. In addition to listing OSA journal articles that cite this paper, citing articles from other participating publishers will also be listed.


« Previous Article  |  Next Article »

OSA is a member of CrossRef.

CrossCheck Deposited