Quantitative Phase Imaging Using A Combination of Flat Fielding and Windowed Fourier Filtering Demodulated by A Graph Cuts Algorithm For Screening Opaque and Transparent Objects
Quantitative Phase Imaging Using A Combination of Flat Fielding and Windowed Fourier Filtering Demodulated by A Graph Cuts Algorithm For Screening Opaque and Transparent Objects
Abstract: In this paper, quantitative phases of opaque and transparent objects are measured
precisely using a combination of flat fielding and windowed Fourier filtering demodulated
by graph cuts algorithm. The modulated interferogram is corrected first by flat fielding to
remove dust particles and adjust inhomogeneity of the interferogram intensity. The corrected
interferogram is then convolved by windowed Fourier filtering to produce an interferogram free
from speckle noise. The obtained interferogram is reconstructed by the Fourier transform method,
or the wavelet method, or the angular spectrum method to extract the wrapping phase of the
object. A graph cuts algorithm is used to unwrap the wrapping phase to remove the 2π ambiguity.
Experimental results show that quantitative phases of the objects being screened are precisely
measured by the proposed method. Moreover, the lateral resolution, which is represented by
slope of the roll-off is slightly improved without applications of digital filters. Furthermore,
shapes of the echinocytes of the cancerous blood cells which have the sharpest spatial features
are seen clearly by the proposed method.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Objects viewed in highly coherent light acquire a peculiar granular appearance [1]. The detailed
structure of this peculiar granular appears chaotic and unordered. When the coherent light is
diffusely reflected from a rough surface with roughness on the scale of an optical wavelength,
the obtained field consists of many wavelets. Interference of the coherent wavelets results in the
granular intensity pattern is defined as speckles. Using mean and median filters [2] are good
examples for speckle noise suppression. However, they do not work well for waving structures in
fringe patterns. A good exception is the sine or cosine filter [3], but it fails to deal with fringe
patterns with fast waving structures [4]. Ibrahim used convolution of windowed Fourier transform
with Chebyshev type 2 and elliptic filters to improve the intensity-contrast image of a noisy digital
hologram [5]. However, enhancement of the phase-contrast image was not considered. Due to
the nature of optical interferometry, the extracted phase may present some ambiguities [6,7].
Such ambiguities are phase-order ambiguity and sign ambiguity. The phase-order ambiguity
can be removed by phase unwrapping, while the sign ambiguity can be mitigated by forcing the
derivatives of local frequency. The high-quality pixels are processed with no error, however,
the low quality pixels may cause an error. In this paper, we proposed a method to increase
the efficiency of low quality pixels to avoid the error in phase-contrast image resulting in the
sign ambiguity. The method is a combination of flat fielding and windowed Fourier filtering
demodulated by graph cuts algorithm. The flat fielding removes dust particles attached with the
interferogram as well as adjusts inhomogeneity of its intensity. The corrected interferogram
is then convolved by windowed Fourier filtering (WFF) to produce an interferogram free from
speckle noise. The obtained interferogram is then reconstructed by the Fourier transform (FT)
#448128 https://doi.org/10.1364/OPTCON.448128
Journal © 2022 Received 10 Nov 2021; revised 4 Jan 2022; accepted 6 Jan 2022; published 3 Feb 2022
Research Article Vol. 1, No. 2 / 15 Feb 2022 / Optics Continuum 247
method, or wavelet transform (WT) method, or angular spectrum method (ASM) to extract
the wrapping phase map of the object being screened. A modified graph cuts algorithm has
been used to unwrap the wrapping phase map to remove the 2π ambiguity [8]. The algorithm
segmented the image into background and foreground images and the procedure continuous for
multiple regions simultaneously until no further segment is found. The proposed method has
been applied to quantify the phase of opaque and transparent objects precisely. For the opaque
objects, off-axis interferograms of large scale step height of 200µm and small scale step height
of 1.50µm were captured and corrected by the proposed method then reconstructed by the FT
method and the WT method. For the transparent objects, inline interferograms were captured
and corrected by the proposed method then reconstructed by the ASM. The purpose for using
inline interferograms in case of transparent objects is to make full use of the space-bandwidth
product (SBP) of the camera and hence can resolve finer spatial details of the biological cells.
Since the conventional reconstructed inline interferogram is always blurred by the transmitted
zeroth-order light and twin image, the proposed method mitigates the zero-order light and twin
image of the inline interferogram significantly. Experimental results show that a systematic
error [9] due to phase ripples, which it depends on the random speckle phase and characterizes
the accuracy of the system [10,11] have been minimized to the least measure. Such ripples
are inevitable due to the side lobes effect when the FT method is used [12,13]. Quantitatively,
over one hundred measurements, the mean standard deviation of the measured error due to
phase ripples has been estimated to be in the range of 0.6rad, that is path length in the range
of 30nm. Since the proposed method minimizes such noise to the least measure, so we claim
that quantitative phase of the object being screened is measured precisely. Moreover, the lateral
resolution, which is represented by slope of the roll-off is slightly improved without applications
of digital filters. Based on the merits of the proposed method, we applied it to extract the phase
map of the shapes of the echinocytes of the cancerous blood cells which have the sharpest spatial
features. To the best of our knowledge, this is the first time that a combination of flat fielding and
WFF demodulated with a graph cuts algorithm is used to precisely quantify the phase-contrast
images.
The FT has poor ability to localize the frequency components of a signal with respect to
position [17]. Windowed-Fourier Transform (WFT) [18,19] has been modified for demodulation
to provide time-frequency representation with better signal localization. WFT is based on division
of the non-stationary signal into small portions, which are assumed to be stationary. This is done
using a window function of a selected width, which is shifted and multiplied with the signal
Research Article Vol. 1, No. 2 / 15 Feb 2022 / Optics Continuum 248
to obtain the small stationary signals. The Fourier transform is then applied to each of these
portions to obtain the required transform of the signal. This WFT method uses a fixed window
length and therefore gives a fixed resolution at all times. The WL method [20] improves the fixed
resolution in the WFT method to some extent. Moreover, it is a suitable method for demodulation
of non-stationary signals as seen in Fig. 1(b) and works well with noisy interferograms with no
necessity for choice of window in the transform plane. The ASM is valid for a shorter distance,
so we used it to reconstruct the modulated interferograms of the transparent objects (biological
blood cells). For the opaque objects, we used the FT method to reconstruct the modulated
interferograms. Since the proposed method based FT may shift the roll-off of the object to the
interior edge (Fig. 5(a)), therefore, we applied the WL method to spatially relocalize the roll-off
of the object to the exterior edge.
where α(x,y), β(x,y) are the background illumination and the amplitude modulation of the fringes,
respectively, fo is the spatial carrier frequency, ϕ(x,y) is the phase modulation of the interferogram
and x, y are axes in x and y, respectively. The expression of Eq. 1 can be re-writing as:
Where C(f - fo , y) and C*(f + fo , y) are the carrier signals and A(f, y) is DC term. Figure 2(a)
shows the one-dimensional (1D) profile along the three spectra in frequency domain, and Fig. 2(b)
shows the selected spectrum in Fig. 2(a) after centering.
A filter window is used to select the carrier signals and shifting that component towards the
origin. The inverse Fourier transform is then applied to the spectrum component C(f, y) to obtain
c(x, y). The real and imaginary components of c(x, y) are expressed as:
The obtained phase is limited between -π/2 and π/2 due to the arctangent function. The
obtained wrapped phase values are then unwrapped by the graph cuts algorithm.
Research Article Vol. 1, No. 2 / 15 Feb 2022 / Optics Continuum 249
Fig. 2. (a) Spectra of a fringe pattern in frequency domain. (b) Selected frequency
component at the center.
where f (x) in our case represents a row of a fringe pattern, * denotes complex conjugation. These
coefficients refer to the closeness of the signal f (x) to the wavelet at a particular scale a and
translation b. The modulus |D(a,b)| and the phase ϕ(a,b) arrays are expressed as
√︂
|D(a, b)| = [Re(D(a, b))]2 + [Im(D(a, b))]2 , (10)
(︃ )︃
Im(D(a, b))
φ(a, b) = tan −1
. (11)
Re(D(a, b))
The obtained wrapped phase values are then unwrapped by the graph cuts algorithm.
Research Article Vol. 1, No. 2 / 15 Feb 2022 / Optics Continuum 250
where the reference beam is assumed to be a plane wave and T (fx , fy ; z) is the spatial frequency
transfer function of propagation through a distance z expressed as [22–24]:
(︃ )︃
2π
√︂
T(fx , fy ; z) = exp j z 1 − (λfx ) − (λfy ) ,
2 2
(13)
λ
√
where j = −1, fx and fy are called spatial frequencies in x and y, λ = 635 nm is the illumination
wavelength, and z is the propagation phase factor. The amplitude and phase of the object have
been calculated directly using MATLAB as (abs (Ψp )) and (angle (Ψp )), respectively. The
obtained wrapped phase values are then unwrapped by the graph cuts algorithm.
3. Experiment
Depending on the sample being investigated, two setup configurations were used. A reflection
configuration type was used for opaque objects (e.g. metallic surfaces), and a transmission
configuration type was used for transparent samples (e.g. biological cells). The experimental
arrangement for reflection and transmission configurations are depicted in Figs. 3(a) and (b),
respectively.
Fig. 3. Mach-Zehnder configurations at (a) reflection and (b) transmission; O, object wave,
R, reference wave; NPBS, non-polarized beam splitter; M1 and M2 , mirrors; OC, objective
condenser lens; MO, microscope objective lens; CCD, charge-coupled device.
NPBS2 . The object is a sample mounted on an xyz-translation stage placed at a distance z from
the hologram plane, whose magnified image is projected on the CCD camera, as well as the
reference beam. In case of transmission configuration (Fig. 3(b)), the collimated beam is divided
into two parts: the reference beam (R) and the object beam (O). After having interacted with the
sample, the object beam is collected by the MO (20x, 0.4NA), recombined with the reference
wave, and the interference is recorded on the CCD camera. Here, the reference lens matches
the curvature induced on the object wave by the MO. The lateral magnification, M = - f OC / f MO ,
where f OC is the focal length of the objective condenser lens, and f MO is the focal length of the
microscope objective. The total magnification can be calculated by the distance between object,
MO and the CCD, or using a standard resolution test version. As seen in both configurations,
no moving parts are presents in the optical design resulting in simplified operations, improved
accuracy, high robustness and real-time imaging with a single acquisition. When the object wave
with intensity I O and a reference wave with intensity I R interfere, the intensity of the interference
pattern recorded on the CCD may be expressed as
√︁
Iint = IO + IR + IO IR cos(φ), (14)
where ϕ is the phase difference between the two waves. This Eq. (14) is exactly the same of Eq. (1)
with α (x, y) IO + IR , β (x, y) (IO IR )0.5 , and fo = 0 (on-axis interferogram). The modulated
interferogram I int (x, y) is then corrected by the flat fielding and the obtained interferogram is Iflat
(x, y) expressed as [25–27]:
W(Iint − Ith )
Iflat = , (15)
(If − Ith )
where W is the average pixel value of the corrected flat field interferogram, I int is the raw
interferogram, Ith is the thermal noise image (Fig. 4(b)), and If is the flat frame image (Fig. 4(a)).
The I flat is then convolved with windowed Fourier transform (WFT) expressed as:
∫+∞∫+∞
WFT(fx , fy ) = Iflat (x, y)r∗ (x, y)dxdy, (16)
−∞ −∞
where r(x,y) exp[-x2 /2σx2 −y2 /2σy2 ] is the Gabor transform, which is a Gaussian function and
the symbol * denotes the complex conjugate operation. Here the σx and σy are the standard
deviations of the Gaussian function in x and y directions, and selected to be 10, respectively.
The Gaussian window is selected because it has higher emphasis on the data near the window
center and lower emphasis on the data further away from the window center. By shifting the
central position of the Gaussian window point by point and computing the WFT of the signal, the
fundamental spectral component of each local signal can be extracted. Then, by integrating all
these spectral components, a universal spectral component can be retrieved as:
∫+∞∫+∞
WFT(f0 , f0 )dxdy = F(f0 , f0 ). (17)
−∞ −∞
Finally, applying the inverse Fourier transform to the universal spectral component, a complex
image is obtained.
The real values of the complex image have been taken to provide a speckle free interferogram.
The obtained speckle free interferogram is then demodulated here by the FT method or by the WL
method or by the ASM to produce wrapped phase map of the object being screened. The wrapped
phase map is then unwrapped by the graph cuts algorithm. In preparation for measurement, all
precautions have been taken into account to minimize the systematic error [10].
Research Article Vol. 1, No. 2 / 15 Feb 2022 / Optics Continuum 252
Fig. 4. (a) Flat frame image captured when the object arm of the interferometer is blocked.
(b) Thermal noise image captured when there is no illumination. The standard deviation of
(a) is calculated to be σ = 11.96. The σ of (b) is calculated to be σ = 1.09. The color bar unit
is (a.u.).
Fig. 5. (a) Original off-axis interferogram of step height of 200µm. (b) Correction of (a)
with WFF. (c) Correction of (a) with flat fielding. (d) Correction of (a) with both WFF and
flat fielding. The interferogram is an image of 512 × 512 pixels.
using Eq. (15). The σ of Fig. 5(c) is calculated to be σ = 25.11. Figure 5(b) shows correction
of Fig. 5(a) with WFF using Eqs. (16) and (17) only without applying flat fielding. The σ of
Fig. 5(b) is calculated to be σ = 17.77. Figure 5(d) shows correction of Fig. 5(a) with combination
of both flat fielding and WFF using Eqs. (15), (16) and (17), respectively. The σ of Fig. 5(d) is
calculated to be σ = 11.22. As seen in Fig. 5(d), the proposed method clears the interferogram
from the coherent noise and increases the SNR of the interferogram by around 30%. To extract
the phase map of the off-axis interferogram, we reconstructed Fig. 5(c) as an example by the FT
method in the absence and in the presence of the object and the difference is then extracted.
Figure 6(a) is the off-axis interferogram in the absence of the object and Fig. 6(b) is its wrapped
phase map. Figure 6(c) is the wrapped phase map of Fig. 5(a). Figure 6(d) is the unwrapped
phase map of Fig. 6(b), while Fig. 6(e) is the unwrapped phase map of Fig. 6(c). Figure 6(f) is
the difference between Fig. 6(d) and Fig. 6(e).
Fig. 6. (a) Off-axis interferogram in the absence of the object. (b) Wrapped phase map (a).
(c) Wrapped phase map in the presence of the object. (d) Unwrapped phase map (b). (e)
Unwrapped phase map of (c). (f) Difference between (d) and (c). The color bar unit is in
radians.
To see the performance of the proposed method, we reconstructed the three off-axis interfer-
ograms of Figs. 5(b), (c), and (d) by the FT method following the same procedure of Fig. 6.
Figures 7(a), (c), and (e) show the three-dimensional (3D) pseudocolor of the reconstructed
phase maps of Figs. (a), (b), and (d), respectively. Figures 7(b), (d), (f) are zoomed in the black
rectangles in Figs. 7(a), (c), (e), respectively. As seen in the black circles of Figs. 7(b) and (d),
some spikes are appeared. Such spikes are disappeared completely in Fig. 7(f). This confirms
that the proposed method, which is a combination of flat fielding and WFF removes the spikes
completely. An interesting observation is that the lateral resolution, which is represented by
slope of the roll-off is slightly improved without applications of digital filters [5].
To see such improvement in lateral resolution, three phase profiles along the red, blue, and
black lines in Figs. 7(b), (d), and (f) were extracted and plotted together as seen in Fig. 8(a). As
seen in Fig. 8(a), the black profile is steeper than the blue profile, which confirms the performance
of the proposed method in enhancing the lateral resolution, which is represented by slope of the
roll-off without applications of digital filters.
Research Article Vol. 1, No. 2 / 15 Feb 2022 / Optics Continuum 254
Fig. 7. (a), (c), (e) are 3D pseudocolor of the reconstructed phase maps of Figs. 5(a), (b),
(d), respectively. (b), (d), (f) are zoomed in the black rectangles in (a), (c), (e), respectively.
The color bar unit is in radians.
Fig. 8. (a)1D phase profiles along the red, blue, and black lines of Figs. 7(b), (d), and (f)
extracted by FT method. (b) 1D phase profiles along the red line of Fig. 7(b) extracted by
the FT method and the WL method.
As seen in Fig. 8(a), the application of flat fielding and WFF makes the roll-off of the object
shifts to the interior edge. This shift (around 20 pixels) is due to the existence of two edges
(interior and exterior) as shown in Fig. 5(a). Without application (W/O) of flat fielding and
WFF the roll-off moves to the exterior edge (red profile of Fig. 8(a)). When we applied the
Morlet wavelet method to check the right location of the profile, we found that the roll-off of the
object moves to the exterior edge. Figure 8(b) shows the 1D phase profiles of Fig. 7(b) without
applications of flat fielding and WFF reconstructed by the FT method (red profile) and by the
Morlet WL method (purple profile). The Morlet wavelet transform method can spatially relocalize
the roll-off of the object to the exterior edge. It is worth mentioning that the lateral resolution is
Research Article Vol. 1, No. 2 / 15 Feb 2022 / Optics Continuum 255
limited by coherent diffraction limit, which is defined by NA of the system and the wavelength
[28]. The opaque object of step height of 1.50µm was positioned gently in the interferometer
and then adjusted carefully to capture the perfect parabolic interferogram. Since we used single
wavelength, so chromatic aberration is nearly negligible [29]. Figure 9(a) shows average of
100 parabolic interferograms with standard deviation σ = 12.8. The corrected interferogram of
Fig. 9(a) by the proposed method is shown in Fig. 9(b) with σ = 5.3. The parabolic interferograms
of Fig. 9(a) and Fig. 9(b) were reconstructed by the FT method [10]. If the original parabolic
interferogram is denoted as Ip (x,y) and the corrected interferogram by the proposed method is
denoted as Ipf (x,y). The Ipf (x,y) is digitized after 2D sampling and the digitized interferogram
Ip (ℵ,l) is reconstructed by computing the Fresnel integral as:
where ψ is the reconstructed wave front, ∆x and ∆y are the sampling intervals in the interferogram
plane, m, n, ℵ, l are integers, λ is the wavelength of the laser light, A is a complex constant
given as A = exp (i2π/λ)/(iπλ), FFT denotes the fast Fourier transform, and ∆ξ and ∆η are the
sampling intervals in the observation plane defined as ∆ξ = ∆η = λd/L. The digital reference
wave RD (ℵ, l) is defined as A exp [(i2π/λ) (kx ℵ∆x + ky l∆y)]. The wave vectors kx and ky are
adjusted such that the propagation direction of RD (ℵ,l) agrees as close as possible with that of
the experimental reference wave. The phase of the simulated digital lens is expressed as φ(m,n)
exp [(-iπ/λD) (m2 ∆ξ 2 + n2 ∆η2) ], where D is expressed as 1/D = 1/d (1 + d0 /d), where d0 is the
distance between the microscope objective and the object, d is the distance between the object
and the CCD camera of size (L). The array of the complex numbers of the conjugate digital lens
is multiplied by the reconstructed wave front to compensate for the wave front curvature. The
result of multiplication is an array of complex numbers denoted as Ψ. The phase contrast image
is obtained using MATLAB as arctan {Re(Ψ)/Im(Ψ)}. The obtained phase is then unwrapped by
the graph cuts algorithm. Figures 9(c) and (d) show the reconstructed phase-contrast images of
Fig. 9(a) and Fig. 9(b), respectively. To see the performance of the proposed in removing the
ripples, phase profiles along the red and blue lines of Figs. 9(c) and (d) were extracted and plotted
together as shown in Fig. 9(e). As seen in Fig. 9(d), the maximum ripple in the red phase profile
is in the range of 0.6 rad, corresponds to a path length of (0.6*635)/ (4*3.14) ≅ 30nm. Such
ripples, which represented by the blue profile in Fig. 9(e), are minimized to the least measure by
the proposed method.
Fig. 9. (a) Original parabolic interferogram of step height of 1.50µm. (b) Correction of (a)
with flat fielding and WFF. (c) Reconstructed phase-contrast image of (a). (d) Reconstructed
phase-contrast image of (b). 1D phase profiles along the red and blue lines of (c) and (d).
kind of technology can be utilized to observe the living cells [32,33]. The refractive index of the
RBCs equals nearly 1.42 [31], so the physical thickness of the cell at the (x, y) position can be
calculated easily once the quantitative phase map ϕ(x, y) at pixels (x, y) is extracted. To extract
the quantitative phase map of the object, ASM was used to reconstruct the inline interferogram.
Unlike the Fresnel diffraction method, the ASM is valid for a shorter distance. For the preparation
of the blood samples, blood from healthy human donor and unhealthy human (with leukemia)
donor were obtained from the blood bank of the University Hospital of the Cairo University,
Cairo, Egypt. The blood cells were displaced on glass substrates for measurements. The blood
samples were investigated by the interferometer whose configuration is shown in Fig. 3(b). For
the two samples, two inline interferograms were captured by the CCD camera and corrected by
the proposed method then reconstructed by the ASM. Figure 10(a) shows off-axis interferogram
(752 × 480) pixels of the blood healthy human donor sample captured when there is a slightly
small angle between the reference and object beams.
Figure 10(b) shows the corrected inline interferogram of (100 × 93) pixels of the red rectangle
of (a) when there is no angle between the reference and object beams. Figure 10(c) shows the
reconstructed phase map of Fig. 10(b) with the ASM. The image plane of the MO is located
approximately 50mm behind the camera. Figure 10(d) shows 3D pseudo color of an individual
cell in the white square of Fig. 10(c) with size of (12 × 12) pixels. As seen in Fig. 10(d), the
normal blood cell has a flattened biconcave disc of about 8µm in diameter with no deformation
in its membrane. The phase profile along the black line of Fig. 10(d) is shown in Fig. 12(a). The
maximum height of the cell at its center is in the range of 2.7 radians. Figure 11(a) shows the
corrected inline interferogram (400 × 420 pixels) of the blood unhealthy human donor captured
when there is no angle between the reference and object beams. Reconstructed amplitude-contrast
and phase-contrast images of Fig. 11(a) are shown in Fig. 11(b) and Fig. 11(c), respectively. The
image plane of the MO is located approximately 22mm behind the camera. Figure 11(d) shows
3D pseudocolor of an individual cell in the red rectangle of Fig. 11(c). As seen in Fig. 11(d), the
membrane of the cell is clearly deformed. The 1D phase profile of Fig. 11(d) along the blue line
of (d) is shown in Fig. 12(b). The maximum height is in the range of 2.7 radians. As seen in
Research Article Vol. 1, No. 2 / 15 Feb 2022 / Optics Continuum 257
Figs. 12(a) and (b), the maximum heights of both the healthy and cancerous cells are congruent.
The main difference between the two cells appears in the deformation of the membrane. The
membrane of the healthy cell appears regular, however, the membrane of the cancerous cell
appears deformed. As seen in Figs. 12(a) and (b), the maximum thickness at its center were
estimated to be in the range of 650nm (using Eq. (19) when the refractive index of the RBCs
equals 1.42 [31]. We claim that the deformation in the cancerous cell of Fig. 11(d) and its profile
of Fig. 12(b) is due to changes in the packing properties of the cell [34,35].
Fig. 10. (a) Off-axis interferogram of size (752 × 480) pixels. (b) Corrected inline
interferogram of the red rectangle of (a) with size of (100 × 93). (c) Reconstructed phase
map of (b) by the ASM with a propagation phase factor z = −50 mm. (d) 3D pseudocolor
of the white rectangle of (c) with size of (12 × 12) pixels. The phases in (c) and (d) are in
radians.
Fig. 11. (a) Inline interferogram of size (400 × 420) pixels. (b) Amplitude-contrast image
of (a) reconstructed by the ASM. (c) Phase-contrast image of (a) reconstructed by the ASM
with a propagation phase factor z = −22 mm. (d) 3D pseudocolor of the selected rectangle of
(c) with size of (100 × 93) pixels. The phases in (c) and (d) are in radians.
Deformation in the cancerous cell membrane may be attributed to changes in the packing
properties of the phospholipid macromolecules forming the RBCs cell wall leading to decreased
Research Article Vol. 1, No. 2 / 15 Feb 2022 / Optics Continuum 258
Fig. 12. 1D phase profiles along the (a) black line of Fig. 10(d) and (b) the blue line of
Fig. 11(d). Blue profile in (b) is the original extracted phase, while the red profile is its
quadratic fitting.
cell membrane elasticity. An interesting observation is viewing bulges in the cancerous cell of
Fig. 11(d).
Since the general shape of the normal blood cells is a discocyte, this gives an evidence that the
screened cell of Fig. 11(d) is cancerous cell and it seems to be echinocyte type I with 9 bulges
(E1-9). Echinocytes in general are characterized by one or more spicules. Echinocyte type I
with 9 bulges (E1-9) is characterized by a 9-fold rotation axis plus up-down symmetry, reflecting
the nine identical bulges that develop on the rim. Types of echinocytes with different bulges are
reported in [36]. The agents of echinocytes shapes may due to cholesterol addition, high salt
(hypertonic saline), high pH, intracellular adenosine triphosphate depletion, or proximity to glass.
5. Conclusion
Phase-contrast images associated with coherent phase noise rendering it unsuitable for quantitative
phase-contrast imaging with high precision. To achieve precise phase measurement, phase noise
should be minimized to the least measure. In this paper, we proposed a method to suppress
the coherent phase noise, especially the ripples to the least measure. The method is based on
a combination of flat fielding and windowed Fourier filtering demodulated by the graph cuts
algorithm. The flat fielding removes dust particles attached with the interferogram as well as
adjusts inhomogeneity of its intensity. The corrected interferogram has been convolved by
the windowed Fourier filtering to produce an image free from coherent noise. The corrected
interferogram is then demodulated by the FT method, or WT method, or ASM to extract the
wrapping phase map of the object being screened. The graph cuts algorithm has been used to
unwrap the wrapping phase map to remove the 2π ambiguity via segmentation of the wrapping
phase image into background and foreground images and the procedure continuous for multiple
regions simultaneously until no further segment is found. We introduced the performance of the
proposed method by screening two opaque objects and two transparent objects. The two opaque
objects were screened by the FT method, while the two transparent objects were screened by the
ASM. Experimental results showed that the proposed method has the ability to minimize the
phase noise, especially the ripples to the least measure. Quantitatively, the measured error due to
ripples has been estimated to be in the range of 30nm. In case of objects having interior and
exterior edges, the proposed method can shift the roll-off of the object to the interior edge. The
Morlet wavelet transform method has been used to spatially relocalize the roll-off of the object to
the exterior edge. By using the proposed method with single shot inline interferogram, we clearly
viewed the bulges, which have the sharpest spatial features in the echinocytes. Conventionally,
the reconstructed inline interferogram is always blurred by the transmitted zeroth-order light and
twin image. By the proposed method, the zero-order and twin image were mitigated drastically.
Moreover, the lateral resolution, which is represented by slope of the roll-off is slightly improved
without applications of digital filters.
Research Article Vol. 1, No. 2 / 15 Feb 2022 / Optics Continuum 259
References
1. P. Jacquot and J. -M. Fournier (Editors), “interferometry in speckle light: theory and applications,” Proceedings of
the international conference 25-28 September, Springer, Lausanne, Switzerland (2000).
2. R. C. Gonzalez and R. E. Woods, Digital Image Processing, 3rd edition, Pearson Education Inc., Upper Saddle River,
pp. 344–357 (2010).
3. H. A. Aebischer and S. Waldner, “A simple and effective method for filtering speckle-interferometric phase fringe
patterns,” Opt. Comm. 162(4-6), 205–210 (1999).
4. Q. Kemao, L. T. H. Nam, L. Feng, and S. H. Soon, “Comparative analysis on some filters for wrapped phase maps,”
Appl. Opt. 46(30), 7412 (2007).
5. D. G. A. Ibrahim, “Improving the intensity-contrast image of a noisy digital hologram by convolution of Chebyshev
type 2 and elliptic filters,” Appl. Opt. 60(13), 3823–3829 (2021).
6. F. Charriere, T. Colomb, F. Montfort, E. Cuche, P. Marquet, and C. Depeursinge, “Shot-noise influence on the
reconstructed phase image signal-to-noise ratio in digital holographic microscopy,” Appl. Opt. 45(29), 7667–7673
(2006).
7. Qian Kemao, “Windowed Fringe Pattern Analysis,” Chapter 3, Spie press, Bellimgham, Washington, USA, (2013).
8. J. M. Bioucas-Dias and G. Valadão, “Phase unwrapping via graph cuts,” IEEE Trans. on Image Process. 16(3),
698–709 (2007).
9. Guide to the Expression of Uncertainty in Measurement. ICS 03.120.30; 17.020 (1996).
10. D. G. A. Ibrahim, “Estimation of an uncertainty budget and performance measurement for a dual-wavelength
Twyman-Green interferometer,” J. Microsc. 282(3), 224–238 (2021).
11. D. G. A. Ibrahim, “Steep Large Film Thickness Measurement with Off-axis Terahertz Digital Holography Recon-
structed by a Direct Fourier and Hermite polynomial,” Appl. Opt. 57(10), 2533–2538 (2018).
12. D. G. A. Ibrahim, “Enhancement of the steepness measurement of a film thickness edge using wavelet transforms
with fringe thinning,” OSA Continuum 3, 1928–1937 (2020).
13. D. G. A. Ibrahim, “Demodulation of a parabolic interferogram in time domain for rough surface characterization,”
Appl. Phys. B: Lasers Opt. 126(9), 146 (2020).
14. D. G. Abdelsalam and D. Kim, “Two-wavelength In-line Phase-shifting Interferometry Based on Polarizing Separation
for Accurate Surface Profiling,” Appl. Opt. 50(33), 6153–6161 (2011).
15. D. G. A. Ibrahim and T. Yasui, “Multi-object investigation using two-wavelength phase-shift interferometry guided
by optical frequency comb,” Appl. Phys. Lett. 112(17), 171101–171104 (2018).
16. Dahi Ghareab Abdelsalam Ibrahim, “Optical Metrology with Interferometry,” Cambridge Scholars Publishing,
London, (2019).
17. J. Zhong and J. Weng, “Spatial carrier-fringe pattern analysis by means of wavelet transform: wavelet transform
profilometry,” Appl. Opt. 43(26), 4993–4998 (2004).
18. Q. Kemao, “Windowed Fourier transform for fringe pattern analysis,” Appl. Opt. 43(13), 2695–2702 (2004).
19. Q. Kemao, “Windowed Fourier transform method for demodulation of carrier fringes,” Opt. Eng. 43(7), 1472–1473
(2004).
20. Qian Kemao, H. Wang, and W. Gao, “Windowed Fourier transform for fringe pattern analysis: Theoretical analyses,”
Appl. Opt. 47(29), 5408–5419 (2008).
21. P. Rastogi and E. Hack, “Phase estimation in optical interferometry”, CRC press, Taylor and Francis group, Boca
Raton, London, New York, (2017).
22. M. Takeda, H. Ina, and S. Kobayashi, “Fourier-transform of fringe-pattern analysis for computer-based topography
and interferometry,” J. Opt. Soc. Am. 72(1), 156 (1982).
23. T.-C. Poon and P. P. Banerjee, Contemporary Optical Image Processing with MATLAB (Elsevier, Oxford, UK, 2001).
24. M. K. Kim, L. Yu, and C. J. Mann, “Interference techniques in digital holography,” J. Opt. A: Pure Appl. Opt. 8(7),
S518–S523 (2006).
25. D. G. Abdelsalam, R. Magnusson, and D. Kim, “Single-shot, Dual-wavelength Digital Holography Based on
Polarizing Separation,” Appl. Opt. 50(19), 3360–3368 (2011).
26. D. G. Abdelsalam and D. Kim, “Coherent noise suppression in digital holography based on flat fielding with apodized
apertures,” Appl. Opt. 19(19), 17951 (2011).
27. D. G. Abdelsalam and D. Kim, “Real-time dual-wavelength digital holographic microscopy based on polarizing
separation,” Opt. Express 285(3), 233–237 (2011).
28. Y. Cotte, F. Toy, P. Jourdain, N. Pavillon, D. Boss, P. Magistretti, P. Marquet, and C. Depeursinge, “Marker-free
phase nanoscopy,” Nat. Photonics. 7(2), 113–117 (2013).
29. D. G. A. Ibrahim, “Simultaneous dual-wavelength digital holographic microscopy with compensation of chromatic
aberration for accurate surface characterization,” Appl. Opt. 58(23), 6388–6395 (2019).
Research Article Vol. 1, No. 2 / 15 Feb 2022 / Optics Continuum 260
30. M. Bardyn, B. Rappaz, K. Jaferzadeh, D. Crettaz, J.-D. Tissot, I. Moon, G. Turcatti, N. Lion, and M. Prudent, “Red
blood cells ageing markers: A multi-parametric analysis,” Blood Transfus. 15, 239–248 (2017).
31. B. Rappaz, A. Barbul, Y. Emery, R. Korenstein, C. Depeursinge, P. J. Magistretti, and P. Marquet, “Comparative study
of human erythrocytes by digital holographic microscopy, confocal microscopy, and impedance volume analyzer,”
Cytometry 73A(10), 895–903 (2008).
32. C. L. Curl, C. J. Bellair, P. J. Harris, B. E. Allman, A. Roberts, K. A. Nugent, and L. M. D. Delbridge, “Quantitative
phase microscopy: A new tool for investigating the structure and function of unstained live cells,” Clin. Exp.
Pharmacol. Physiol. 31(12), 896–901 (2004).
33. S. Bhatt, A. Butola, S. R. Kanade, and A. Kumar, “High-resolution single-shot phase-shifting interference microscopy
using deep neural network for quantitative phase imaging of biological samples,” J. Biophotonics 14(7), 1–9 (2021).
34. D. Kabaso, E. Gongadze, S. Perutkova, C. Matschegewski, V. Karalj-Iglic, U. Beck, U. Van Rienen, and A. Iglic,
“Mechanics and electrostatics of the interactions between osteoblasts and titanium surface,” Comput. Methods
Biomechanics Biomed. Eng. 14(5), 469–482 (2011).
35. A. G. Petrov, “Flexoelectricity of model and living membranes,” Biochim. Biophys. Acta 1561(1), 1–25 (2002).
36. G. Gompper and M. Schick, “Soft Matter: Lipid Bilayers and Red Blood Cells,” WILEY-VCH Verlag GmbH & Co.
KGaA, Weinheim, (2008).