art ic l e i nf o
a b s t r a c t
Article history:
Received 1 April 2013
Received in revised form
9 July 2013
Accepted 10 July 2013
Available online 10 August 2013
A multi-level spiral phase lter is proposed in electronic speckle pattern interferometry (ESPI) for out-ofplane displacement measurements. This lter generates a particular kind of speckle pattern that results
from the convolution of standard speckles with the lter point spread function (Fourier transform).
We shall refer to it as a vortex-ltered speckle pattern (VF-SP). It is shown here that if unresolved and
fully developed VF-SPs are generated, then each speckle contains embedded phase terms which can be
controlled by the multi-level spiral lter rotation. This mechanism effectively allows the application of
standard phase extraction procedures for displacement measurements. Numerical simulations of an
interferometer working with VF-SPs were done to verify this technique. Experimental validation was
achieved with an out-of-plane electronic speckle pattern interferometer, in which an Liquid Crystal on
Silicon (LCoS) was used to generate the multi-level spiral phase lters.
& 2013 Elsevier Ltd. All rights reserved.
Keywords:
Speckle
Phase measurement
Vortices
Multi-level spiral phase lters
Spatial light modulators
1. Introduction
Speckle interferometry is a technique for displacement measurements that uses rough surfaces illuminated by laser light,
assuming that such surfaces are capable of generating fully
developed speckle patterns [1]. Several interferometric techniques
for displacement measurements have been devised for various
sensitivities and applications, including in-plane, out-of-plane and
shearing methods [2]. Their arrangements include an iris diaphragm in the pupil plane of the system to control the speckle
size. Further developments of these techniques have recently
added 4f optical correlator systems for various purposes. These
include increasing the eld of view of speckle shearing interferometers [3,4], and introducing tilts and shears in the pupil plane
with a spatial light modulator (SLM), allowing thus the compensation of rigid movements of the object under test [5]. In all of these
cases, the imaging and eld lenses were placed in front of the 4f
system. Knowing this, what we propose here is using a 4f optical
correlator system, but for yet another end. We will show that
phase stepping can be attained in out-of-plane speckle interferometry by spatial ltering. For this end, we will use a multi-level
vortex spatial lter oriented at 01, 901, 1801 and 2701. To our
knowledge, the use of vortex spatial lters is presently unexplored
in speckle interferometry applications involving phase stepping.
Corresponding author. Tel.: +52 477 4414200/201; fax: +52 477 4414209.
E-mail addresses: abundio.davila@me.com, adavila@cio.mx (A. Dvila).
0143-8166/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.optlaseng.2013.07.007
Given the random nature of the speckle phase, there are chances
that optical vortices occur naturally in speckle interferometry, and its
density has been found to depend on the roughness of the samples
[6]. Tracking of the vortex positions has allowed applications for
nanometric displacement measurement [7] and dynamic speckle
from biological samples [8]. Recently, continuous phase vortices have
been used as spatial lters for phase contrast microscopy [9,10].
Microscopic specimens are typically smooth, so that they can be
regarded as weak diffusers, producing undeveloped speckle patterns,
that is, speckle patterns where the random phase excursion of the
eld is {2. When a vortex-ltered, undeveloped speckle pattern is
combined with a smooth reference wave, it yields an interference
pattern with spiral fringes, and phase extraction becomes an involved
task, specic for this kind of fringe patterns [10]. Fortunately, in ESPI
we deal with objects which are optically rough (random phase
excursions of the eld 2, producing fully developed speckle
patterns. This is also valid for VF-SPs, which means that we can
apply to them the usual phase extraction algorithms of ESPI, as we
shall prove in Section 2.
Continuous phase vortices, like those used in microscopy, are
difcult to generate and unsuitable for high speed speckle correlation interferometry. Therefore, a multi-level spiral phase vortex is
proposed here to solve these problems, and also to reduce the
complexity of the phase extraction process. Our aim was to process
fully developed speckle patterns of an out-of-plane speckle interferometer which includes a multi-level spiral phase lter, and
nding ways of extracting the juxtaposed phase in the VF-SP. Since
the technique requires only one phase vortex in the Fourier
domain, we assumed that a modication of this vortex would
20
1
is the object complex amplitude distribution, Pxo ; yo is the generalized pupil function, and
Hn u; v circ =0 exp i n u; v;
3
where
Hn u; v Hn
xf y f
;
f f
4
HuHv;
HuHv;
HuHv T ;
and
R r i;j
r1 ; r2 ; r3 ; r4 T
0
1
0 1 2 3
B3 0 1 2C
B
C
B
C
@2 3 0 1A
1
Fig. 1. Schematic diagram of a 4f optical correlator with a phase vortex lter represented in gray levels.
21
h
i
exp in1 H1 u; v;
2
11
given that
4
j1
HuHv 1:
12
r exp in1 1 H1 u; vGf u; v 2
2i
h
r exp in1 U i1 xi ; yi 2 :
2
Developing the last expression we obtain
I n xi ; yi axi ; yi bxi ; yi cos i xi ; yi n1 =2
13
14
where
axi ; yi r 2 jU i1 xi ; yi j2 ;
Fig. 2. Phase delays of the four quadrants of the multi-level spiral phase vortex in
the four orientations of the major phase discontinuity (03=2) that were used for
spatial ltering (phase stepping) within the speckle interferometer. In the 3D
schematic representations of the lter the thickness is proportional to the phase
delay, which for greater clarity has also been coded in gray levels.
bxi ; yi 2rjU 1
i xi ; yi j;
i xi ; yi argU i1 xi ; yi :
15
I xi ; yi jr
2
U n
i xi ; yi j
jr
1
fH u; vGf u; vgj :
Since
H u; v circ =0 expi n u; v
"
#
4
circ =0 exp i r n;j qj u; v
2j1
16
therefore
tan i xi ; yi
I 4 xi ; yi I 2 xi ; yi
I 1 xi ; yi I 3 xi ; yi
zxi ; yi :
17
18
where
o xo ; yo ~ o xo ; yo o xo ; yo
19
1
n
20
I~ xi ; yi r exp in1 =2 U~ i xi ; yi 2
where
1
U~ i xi ; yi 1 fH1 u; vfPxo ; yo U~ o xo ; yo gg
mod4;
"
#
4
expi 1 exp in1 qj u; v
0
2j1
and
Hn u; v circ
21
2
10
22
In other words, the phase increments introduced by the displacement of the object at the object and the image elds are
approximately equal at conjugate points xi ; yi Mx0 ; My0
x0 ; y0 , where M is the magnication of the system, which
is 1 in the case of a 4f correlator. Therefore
1
~ ; y cos ~ x ; y ;
~ i ; yi bx
I~ xi ; yi ax
i i
i i i
22
2
~ ; y sin ~ x ; y ;
~ i ; yi bx
I~ xi ; yi ax
i i
i i i
33
~ ; y cos ~ x ; y ;
~ i ; yi bx
I~ xi ; yi ax
i i
i i i
4
~ ; y sin ~ x ; y ;
~ i ; yi bx
I~ xi ; yi ax
i i
i i i
23
with
1
~ i ; yi r 2 jU~ i xi ; yi j2 axi ; yi ;
ax
tan i xi ; yi
~ ; y 2rjU~ 1 x ; y j bx ; y ;
bx
i i
i i
i i
i
1
~ i xi ; yi argU~ i xi ; yi i xi ; yi i xi ; yi :
24
4
2
I~ xi ; yi I~ xi ; yi
z~ xi ; yi ;
1
3
I~ xi ; y I~ xi ; y
i
25
26
27
29
where
i xi ; yi 2i xi ; yi i xi ; yi :
30
Let
f x; yAi
Ai f x; ydx dy
Ai dx dy
31
32
From the averages of the two differences [Eq. (32)], and taking
into account that averaging is a linear operation [Eq. (31)], we
obtain
4 x; yAi 2 x; yAi
;
1 x; yAi 3 x; yAi
34
which is similar to Eq. (25), except that it gives directly the value of
i xi ; yi . After this, e
o xo ; yo can again be readily approximated with Eq. (27).
Similar equations for phase extraction can be found in the case
of continuous phase vortices, and also of phase vortices with
topological charges different from onewith phase spans larger
than 2. At any rate, with the method that we propose here, the
phase steps that are required for phase extraction can be accurately controlled by the vortex orientation, without the need of an
elaborate calibration procedure to avoid detuning errorslike in
the case of PZTs. The maximum speed for the rotation of the vortex
depends on the device that we use to produce it. New, non-LC,
SLMs have been recently introduced to increase the video rate
speeds of current LC devices, with an almost instantaneous
response [19].
3. Numerical simulations
3.1. Numerical verication of the phase extraction equations
To verify the phase extraction equations, Eqs. (26) and (34), we
devised a numerical simulation of an out-of-plane ESPI system
working with VF-SPs. Preliminary values for the amplitude
Ao xo ; yo and phase o xo ; yo of the object eld were obtained
with a random number generator. These values were subsequently
ltered with a low-pass lter, so that the speckle size of the eld
U o xo ; yo was adjusted through the cut-off spatial frequency of
the lter. The eld U i xi ; yi at the image plane of the 4f correlator
(Fig. 1) was obtained after two numerical Fourier transformations,
indicated in Eqs. (1) and (3). After computing this eld, the
intensity resulting from the superposition of U in xi ; yi and a
plane wave of constant amplitude r and zero phase is found. As a
result of this numerical simulation, the intensity distributions
n
I n xi ; yi and I~ xi ; yi were obtained, in the second case after
adding a Gaussian phase increment o xo ; yo 2h=exp
x2o y2o =w2 to the object phaseo xo ; yo , with w 5 mm, and
h 4. The object area was 10 10 mm2.
Fig. 3 shows the phase map that was extracted with an
algorithm based on Eq. (26), for the Gaussian phase increment
mentioned above. The map shown has been smoothed with a
3 3 median lter. We can notice that the resemblance in shape,
height h= 4 and width w 5 with the original phase
increment map is quite close.
The drawback of an algorithm based on Eq. (26) is that we need
to record eight intensity patterns, four before the displacement of
the object, and four after. This number can be reduced to ve if the
phase increment is extracted from correlation fringe patterns, that
is, with an algorithm based on Eq. (34). Fig. 4 shows a set of these
patterns. After smoothing them with a low pass lter, we obtained
the phase map shown in Fig. 5. It is similar to the map of Fig. 3,
except that it still keeps some speckle noise. To remove it, a broad
range of lters have been proposed: from simple ltering, such as
neighborhood average, up to scale-space and regularization lters
[20,21].
To test the reliability of both methods, we made use of the
Universal Image Quality Index Q, which provides a measure of
23
Fig. 3. Phase maps extracted by using Eq. (26). (a) Wrapped phase map and (b) unwrapped phase map.
Fig. 4. Correlation fringe patterns obtained with the numerical simulation: (a) 1 xi ; yi ; (b) 2 xi ; yi ; (c) 3 xi ; yi ; (d) 4 xi ; yi .
Fig. 5. Extracted phase obtained from the correlation fringe patterns of Fig. 4 and an algorithm based on Eq. (34).
Fig. 6. Interferograms of the LCoS obtained with a Fizeau interferometer (a), and generated after a least squares t of a low order polynomial to the OPD values provided by
the interferometer (b). The bright areas show the actual size that was given to the phase vortex lter with a diaphragm.
35
24
increment e
o xo ; yo , the quality indices that we obtained when
the phase was extracted with algorithms based on Eqs. (26) and
(34) were Q 0.97 and Q0.95, respectively, showing that both
methods are equally reliable.
3.2. Effect of the aberrations of the phase lter
The phase vortices that we used were generated by a reective
spatial light modulator of the liquid crystal on silicon type (LCoS).
Now, it is well known that an active LCoS introduces an spurious
phase throughout its surface, which corresponds to a wavefront
aberration of several wavelengths [23]. This aberration is difcult
to compensate with the LCoS itself, because the LCoS phase
modulation depth is at the most 3, that is, the equivalent of an
optical path difference (OPD) of only 1.5 wavelengths.
The effect of aberrations of phase vortices in spatial ltering
has been studied in detail by Sharma and Senthilkumaran [24]. We
shall simulate here their effect in the case of speckle interferometry with VF-SPs, to see the extent to which the aberrations
affect the phase which is extracted from the correlation fringes.
For this end, a phase term needs to be added to the phase n u; v
of the vortex, to take into account the wavefront aberration
Table 1
Coefcients of the low order polynomial that we used to represent
the wavefront aberration function of our LCoS.
Coefcient
A
B
C
D
E
F
1.85
0.59
2.85
0.04
0.05
1.23
Hn u; v exp i n u; v 2= Wf u; f v :
36
For moderate aberrations, Wxf ; yf can be adequately represented by a low order polynomial
!2
!2
yf
yf
xf
xf yf
xf
Wxf ; yf = A
B
C
D E F;
37
lxf
lxf lyf
lyf
lxf
lyf
where 2lxf and 2lyf are, respectively, the width and the height of
the LCoS. Notice that the coefcients A, B, C, D, E and F in Eq. (37)
are adimensional.
Fig. 6a shows an interferogram obtained with our LCoS in a
Wyko interferometer. From the raw data of the interferometer
(the OPDs given in micrometers for each pixel of the interferogram), we adjusted the low-order polynomial to the wavefront
aberration measured in wavelengths ( 0:6328 mm). Normalized
pixel numbers were used to assign normalized coordinates xf =lf
and yf =lf to the values of Wxf ; yf =. Table 1 shows the coefcients
that we obtained. Fig. 6b shows the interferogram generated with
the coefcients in this table.
With Eq. (37) and the values of the coefcients in Table 1 we
computed the correlation fringe patterns shown in Fig. 7, which
clearly show a reduced contrast after the lter rotation. Nonetheless, with the values of n xi ; yi that we used to nd these
patterns and Eq. (34) we could nd the phase map in Fig. 8, that is
close enough to the map that we obtained with the phase vortex
that was free from aberrations (Fig. 5).
4. Experimental results
Fig. 9 shows the experimental setup for the out-of-plane deformation measurements. The light from a NdYag laser L 532 nmis
Fig. 7. Progressive degradation of correlation fringes due to phase vortex aberrations. The circles on the lower right corner of each pattern show the lter orientation that
was used to generate the pattern.
Fig. 8. Extracted phase in the case of a moderately aberrated phase vortex lter. It was obtained from the correlation fringe patterns of Fig. 7 and an algorithm based on Eq. (34).
25
divided by a beam splitter BS1 into an object beam, shown with a solid
line, and a reference beam, shown with a dashed line. The object beam
is expanded with a microscope objective MO1 to illuminate an
aluminum disk AD. The light scattered by the disk surface passes
through a polarizer P1 and a half-wave plate HWP, which allows the
adjustment of the plane of polarization of the object beam, before this
is incident on the Liquid Crystal on Silicon (LCoS) spatial light
modulator (SLM). The aluminum disk is placed at the back focal plane
of a lens L1, with 400 mm focal length and 35 mm diameter. The LCoS,
model LC-R 2500 by Holoeye, and a 19 mm diameter diaphragm D
are placed at the back focal plane of L1, which also is the front focal
plane of the lens L2, a lens with the same focal length and diameter
than L1.
The light reected from the LCoS is brought to L2 by means of a
non-polarizing beam splitter BS2. The reference beam generated
by the beam splitter BS1, on the other hand, is directed by the
mirrors M1, M2 and M3 to the microscope objective MO2, also at
the back focal plane of L2. This lens collimates the beam, which is
incident normally at the CCD. For this work, the eld of view of the
spatial ltering system was restricted to that of a unit magnication, 4f system, so that the inspection area was approximately the
size of the CCD. However, this limitation can be overcome, as
shown, for instance, in [3,4] for the case of shearography.
After L2, the object beam and the reference beams present
different states of polarization, reducing thus the contrast of the
correlation fringes. To improve this contrast, a quarter wave plate
QWP and a polarizer P2 were included before the CCD camera.
Trough rotations of QWP and P2 we could optimize the contrast of
the correlation fringe pattern.
The aluminum disk AD was rmly clamped at its rim, and
deformed with a pointed rod PR. The rod was xed to a translation
stage equipped with a micrometer head, where we could accurately
read the displacement of the rod. Fig. 10 shows consecutive correlation
fringe patterns obtained after the plate had been deformed, one
pattern for each of the four vortex orientations displayed at the LCoS
(Fig. 2). A slight decorrelation is observed on the fringe patterns that
can be attributed to the phase vortex aberrations (Section 3) and to the
icker of the LCoS, which induces phase uctuations within a band
that can be as large as 0:4721. This band, however, could be
substantially reduced by cooling down the LCoS, as shown in [25].
Fig. 11 shows in gray levels the actual displacement map that
we obtained from the correlation fringes of Fig. 10. We can notice
that the maximum displacement was approximately 2.4 mm,
which agrees well with the number of correlation fringes in
Fig. 10, and the displacement that we gave to the pointed rod.
Fig. 11. Contour map of disk displacements obtained from the correlation fringes is
shown in Fig. 10. The contour intervals are of 0.25 mm.
5. Conclusions
We presented an alternative method of phase extraction from
fringe correlation patterns obtained with vortex-ltered speckle
patterns (VF-SP). We showed that these patterns can be generated
by accurate phase stepping, which is achieved by three consecutive 901 turns of a four-step spiral phase vortex.
The main advantage of this method of phase extraction is that an
accurate phase stepping can be achieved easily, as the accuracy of the
phase steps depends only on the precise rotation of the vortex lter,
and this can be attained with ease using a spatial light modulator. Our
technique is, therefore, virtually immune to detuning problems,
unlike the case of PZTs, where hysteresis and wobbling must be
taken into account. Even though the proposed setup has more
components than the traditional out-of-plane interferometer arrangement, once the interferometer is built, the frequent phase recalibration work of traditional ESPI is not required. In addition, the
incorporation of very fast spatial light modulators into the electrooptics market in the near future will allow our method good control
of phase stepping at very high speeds.
We gave a theoretical justication of this method, and of a
method based on the recording of eight VF-SP, one for each of the
four vortex lter orientations, before and after the object is
deformed. The equations derived from this analysis were experimentally tested, and also veried by numerical simulation.
Although aberrations in the phase vortex spatial lter affect the
visibility of the correlation fringes, from the degraded fringe
patterns we can still extract faithful phase information, since the
Fig. 10. Four-steps experimental correlation fringes: (a) 1 xi ; yi ; (b) 2 xi ; yi ; (c) 3 xi ; yi ; (d) 4 xi ; yi . The inspected area was 10 6 mm2.
26