2. L. Huang et al. Optics and Lasers in Engineering 107 (2018) 247–257
Fig. 1. The measurement principle of the PMD is based on the law of reflection.
The reflected ray doubles the angle change introduced by the specimen slopes.
Fig. 2. For the same camera probe ray and its corresponding phase point on the
screen, there are many possible solutions with different height-slope combina-
tions.
to make the camera(s) see the fringe patterns on screen(s) via the re-
flection by the SUT.
(2) Capture the fringe images of the reflected pattern from the SUT.
(3) Analyze the fringe patterns to retrieve two-directional 2D phase in-
formation, and furthermore, the x- and y-slope values.
(4) Reconstruct the height information from slope datasets which is also
known as the 2D integration process.
2.1. Setups of PMD
The main devices for a general PMD include a digital camera, a TFT
LCD monitor, a specimen stage, and a computer. Computer-generated
fringe patterns are sequentially displayed on the TFT LCD screen. The
camera captures the reflection images of the displayed fringe patterns
via a specular reflecting SUT. The SUT shape is reconstructed by solving
an inverse problem with the captured images. Owing to the height-slope
ambiguity described in Fig. 2, several different kinds of regularization
approaches with the corresponding setups are proposed to resolve this
ill-posed problem.
To measure micron-level out-of-plane deformation or discrepancy
from a pre-known shape, the monoscopic PMD illustrated in Fig. 3 is a
simple and effective solution with the initial shape regularization [25],
which assumes the shape after deformation is very close to the reference
[35, 36], or the SUT is very similar to the pre-known shape [37]. Self-
consistent height and slope results can be obtained via iterations [34].
For different requirements and corresponding regularization in ac-
tual measurements, the PMD system may adopt additional components
(e.g. screens, distance sensors, or cameras). By introducing additional
patterns in the optical path as shown in Fig. 4(a), the reflected ray can
be determined by the two or more points of intersection on the shifted
screen [26, 38–40]. In Fig. 4(b), an additional distance sensor can be
used to regularize the ill-posed problem in the monoscopic PMD by pro-
viding a reference distance [41]. Fig. 4(c) and (d) describe the height
values can be searched by minimizing the discrepancies of the SUT nor-
mal vectors calculated from two or multiple cameras [24]. These cam-
eras can be served by a single screen as Fig. 4(c) or several screens
as Fig. 4(d). In addition, some recently proposed new configurations
require other constraints, such as parallel screens and reference plane
[42].
2.2. Image acquisition
In PMD measurement, a camera captures the reflection of the pat-
terns displayed on a screen through the specimen surface. There is a
trade-off between the spatial resolution and the angular resolution in
image acquisition. If the camera is focused on the specimen surface, the
measurement gets the best spatial resolution, but the angular resolution
will not be the optimum due to the defocusing of the screen patterns. If
the camera is focused on the reflection of the screen pattern, the mea-
surement achieves the best angular resolution but a lower spatial resolu-
tion comparing to the previous case. In practice, the cameras are usually
focused on the SUT for the following practical considerations.
(1) High spatial resolution is a common requirement for 3D shape mea-
surement, if achievable.
(2) The patterns displayed on the screen in PMD are typically sinusoidal
fringes which are smooth intensity curves and the phase calculation
is not very sensitive to a small amount of out-of-focus effect;
(3) There is less influence from the pixel grids of TFT LCD screen. Owing
to the defocusing effect, the TFT LCD pixel grids almost disappear in
the camera image, which is preferable. On the other hand, when the
camera focuses on the reflection of the screen pattern, the images
will record the TFT LCD pixel grids, which introduces an additional
error source to the follow-up fringe analysis process.
2.3. Fringe analysis and slope calculation
Once the fringe patterns are captured, the fringe phases need to be
retrieved by using the well-developed fringe analysis method as one of
the most significant intermediate results. The fringe analysis includes
fringe demodulation and phase unwrapping.
2.3.1. Fringe demodulation
Fringe demodulation allows retrieving the wrapped phase values
from the captured fringe intensity image(s). According to the required
248
3. L. Huang et al. Optics and Lasers in Engineering 107 (2018) 247–257
Fig. 3. The basic monoscopic PMD simply consists of a screen and a camera, which displays and captures fringe patterns, respectively.
Fig. 4. Some other typical PMD setups: (a) monoscopic PMD with shifted screens, (b) monoscopic PMD with a point distance sensor, (c) stereoscopic PMD, and (d)
multi-camera PMD with several screens serving different cameras.
Fig. 5. Since two-directional phases are usually needed in PMD, either a crossed fringe pattern (a) or one-directional fringe patterns in the x-direction (b) and
y-direction (c) are displayed on the screen and captured by the camera(s).
number of frames, they can be classified as single-frame and multiple-
frame methods.
(1) Single-frame methods need a single fringe pattern to calculate
the phase information. They are mainly transform-based methods
[43], such as windowed Fourier transform [21, 22, 35, 36]. As
two-directional phase values are required to calculate both x-and y-
slopes, a crossed fringe pattern as shown in Fig. 5(a) is used to carry
the two-directional phases. The captured crossed fringe intensity I(x,
y) can be expressed as
𝐼(𝑥, 𝑦) = 𝑎(𝑥, 𝑦) + 𝑏𝑥(𝑥, 𝑦) ⋅ cos
[
𝜙𝑥(𝑥, 𝑦)
]
+ 𝑏𝑦 ⋅ cos
[
𝜙𝑦(𝑥, 𝑦)
]
, (1)
where x and y are the orthogonal coordinates of the screen, a(x, y) is
the background, bx(x, y) and by(x, y) are the modulations of x- and y-
249
4. L. Huang et al. Optics and Lasers in Engineering 107 (2018) 247–257
Fig. 6. Wrapped phases (a) and (b) need to be unwrapped to the absolute phase values in x- (c) and y-directions (d) for the following slope calculations in PMD.
directional sinusoidal fringes. 𝜙x(x, y) and 𝜙y(x, y) are the fringe phases
in x- and y-directions. By analyzing or filtering in the frequency domain,
the two wrapped fringe phases 𝜙𝑤
𝑥 (𝑥, 𝑦) and 𝜙𝑤
𝑦 (𝑥, 𝑦) can be retrieved as
𝜙𝑤
𝑥 (𝑥, 𝑦) = arctan
Im
[
̄
𝑓𝑥(𝑥, 𝑦)
]
Re
[
̄
𝑓𝑥(𝑥, 𝑦)
] , (2)
𝜙𝑤
𝑦 (𝑥, 𝑦) = arctan
Im
[
̄
𝑓𝑦(𝑥, 𝑦)
]
Re
[
̄
𝑓𝑦(𝑥, 𝑦)
] , (3)
where ̄
𝑓𝑥(𝑥, 𝑦) and ̄
𝑓𝑦(𝑥, 𝑦) denote the filtered x- and y-directional expo-
nential fringe patterns.
(2) Multiple-frame methods use several phase shifted fringe images to
estimate the phase values in the least squares sense, e.g. the well-
known phase shifting or say phase stepping methods [18, 20]. Usu-
ally, the phase shifted fringe pattern In(x, y) is in a one-dimensional
sinusoidal waveform as
𝐼𝑛(𝑥, 𝑦) = 𝑎(𝑥, 𝑦) + 𝑏(𝑥, 𝑦) ⋅ cos
[
𝜙(𝑥, 𝑦) +
2𝑛𝜋
𝑁
]
, 𝑛 = 0, 1, 2, … , 𝑁 − 1.
(4)
where a(x, y), b(x, y), and 𝜙(x, y) are the background, modulation, and
phase of x- or y-direction fringes, respectively. The wrapped phase 𝜙w(x,
y) can be calculated as
𝜙𝑤
(𝑥, 𝑦) = − arctan
𝑁−1
∑
𝑛=0
𝐼𝑛 sin
(
2𝑛𝜋
𝑁
)
𝑁−1
∑
𝑛=0
𝐼𝑛 cos
(
2𝑛𝜋
𝑁
)
. (5)
Worthy to note, unlike the usual approaches, Ref. [44] introduced
a novel fringe demodulation method to calculate the two-directional
phase information of crossed fringe patterns with phase shifts along one
direction.
After the fringe demodulation, the phase values are wrapped within
[-𝜋, 𝜋] as shown in Fig. 6(a) and (b) because of the four-quadrant inverse
tangent function in Eqs. (2), (3) and (5). In order to calculate slopes
by using phase values, these wrapped phases need to be unwrapped to
absolute phases.
2.3.2. Phase unwrapping
Phase unwrapping [45] extends the wrapped phase values to break
the limit of [-𝜋, 𝜋] as illustrated in Fig. 6. According to the unwrapping
domain, the phase unwrapping can be divided into the spatial phase un-
wrapping [46, 47] and temporal phase unwrapping [48–50]. The spatial
phase unwrapping determines the fringe orders based on the phase re-
lations between neighboring pixels in space, and the temporal phase
unwrapping makes pixel-independent calculation of the fringe orders
along the time axis.
The required phase values in PMD are absolute phases, which means
the fringe orders should be consistent for all measurements. Marker-
assisted spatial phase unwrapping and multi-frequency temporal phase
unwrapping are feasible to calculate the absolute phases in practice:
(1) Marker-assisted spatial phase unwrapping. The specular speci-
mens are usually smooth and continuous surfaces, such as mirrors.
The noise on phase values are generally low and the phase is easy
to unwrap in most cases. The spatial phase unwrapping methods
[46, 47] are good solutions for such a simple unwrapping task, but
additional markers may be needed to determine the reference fringe
orders to keep the consistency of fringe orders in each measurement.
(2) Multi-frequency temporal phase unwrapping. By sequentially
displaying and capturing additional fringe patterns with several de-
signed frequencies, the multi-frequency temporal phase unwrapping
methods [48–50] utilize the phase relations between frequencies to
250
5. L. Huang et al. Optics and Lasers in Engineering 107 (2018) 247–257
Fig. 7. Surface normal N can be calculated from the normalized vector of probe
ray p and reflected ray r.
determine the fringe orders. Comparing to the spatial phase unwrap-
ping, it is a more general solution for the absolute phase calculation.
In most of PMD applications, multi-frequency temporal phase un-
wrapping is employed as a flexible and robust phase unwrapping
solution.
2.3.3. Slope calculation
Once the absolute phase values are retrieved, their locations on the
screen can be determined since the period of the fringe pattern on the
screen is a known parameter, if the pattern is calibrated or the screen
pixel size can be trusted. Moreover, with the geometric calibration pa-
rameters, the global coordinates of the phase-marked intersection point
m in Fig. 7 can be determined. Furthermore, the normalized vector of
the reflected ray r can then be calculated with one of the regularization
approaches according to setups in Figs. 3 and 4.
For each pixel, the normalized vector of the probe ray p is deter-
mined by the camera calibration [51, 52]. As illustrated in Fig. 7, the
SUT normal vector N can be determined by
𝐍 = 𝐫 − 𝐩 =∶
⎛
⎜
⎜
⎝
𝑁𝑥
𝑁𝑦
𝑁𝑧
⎞
⎟
⎟
⎠
, (6)
where Nx, Ny, and Nz are the x-, y-, and z-components of the surface
normal N. The specimen surface x- and y-slopes (sx, sy) are therefore
calculated as
𝑠𝑥 = −
𝑁𝑥
𝑁𝑧
, (7)
𝑠𝑦 = −
𝑁𝑦
𝑁𝑧
. (8)
Once the slopes (sx, sy) and in-plane coordinates (x, y) are calculated,
the shape can be reconstructed from these gradient data.
2.4. Shape reconstruction from slopes
As shown in Fig. 8, the height distribution z is reconstructed from
the calculated coordinates (x, y) and slopes (sx, sy). This 2D integration
process can be express as
𝑧 = 𝑓int2
(
𝑥, 𝑦, 𝑠𝑥, 𝑠𝑦
)
, (9)
where 𝑓int2(⋅) stands for a 2D integration function.
Of course, if there are other trustable information, they can be in-
cluded in the integration process as constraints to obtain a more reliable
optimization. There are mainly three classes of reconstruction methods:
(1) Zonal reconstruction. From the Fried’s, Hudgin’s, and Southwell’s
algorithms in late 1970 s and early 1980 s [53–56] to the recent de-
velopment and improvement especially for PMD technique [57–62],
these zonal methods use the finite differences of height along x-and
y-directions to establish the relations between the measured slopes
with the unknown height values. Iterative compensations [57],
higher order difference format [58], radial-basis-functions-based as-
sistance [60], splines [61], and quadrilateral geometry [61] are
investigated in order to accurately estimate the height values by
matching their derived first derivatives with the measured slopes
in the least squares sense.
(2) Modal reconstruction. These reconstruction methods are based on
analytical models. By taking the first derivatives in x- and y-direction
of the analytical expressions, the model coefficients can be approx-
imated by fitting the measured slopes with analytical slopes. The
commonly used models include polynomials (e.g. Zernike [63–66],
Chebyshev [66], Legendre [67] and B-spline [64, 68]) and sinusoidal
curves (e.g. Fourier transform [69] and cosine transform [70]). Once,
the model coefficients are estimated from the slope fitting, the height
distribution can be calculated by using the coefficients as a weight
onto their corresponding modes.
(3) Piecewise reconstruction. The shape reconstruction method by us-
ing radial basis functions [71, 72] works subset by subset (usually
40 × 40 pixels or less) due to its large memory cost. A stitching pro-
cess is required if the dataset is larger than a single subset. A zonal
method can assist in stitching the height pieces [73].
3. Background and recent development
As the pioneers in profiling specular surface, Petz et al. conducted the
studies on measuring specular reflecting surface by imaging of two grat-
Fig. 8. The x-slope (a) and y-slope (b) can be used to reconstruct the height (c) via a 2D integration process.
251
6. L. Huang et al. Optics and Lasers in Engineering 107 (2018) 247–257
ing planes which were physically moved in parallel [26, 27]. The year
of 2004 was a big year for PMD technique. Knauer et al. [24] published
their work on measuring specular free-form surfaces with deflectometry
based on fringe phase measurement, named phase measuring deflec-
tometry for the first time. In this work, many technical aspects were
addressed, including the fundamental principle, physical limit, system
calibration, and stereo solution for ambiguity. Bothe et al. [25] pre-
sented their research on specular surface measurement by using fringe
reflection technique from the phenomena of the reflection to the princi-
ple of the technique, and to the practical demonstration. Since then, as
a novel technique for free-form specular shape measurement, PMD has
been widely spread and studied by many research groups for different
applications.
(1) Optical inspection. Häusler et al. proposed a microscopic PMD sys-
tem with nanometer sensitivity for local surface features [74]. Tang
et al. utilized their modified PMD method to measure the 3D shape
of aspheric mirrors [37]. Balzer and Werling gave an excellent re-
view on the shape measurement from specular reflection from prob-
lem modeling to hardware and strategies selections [28]. Faber et al.
compared several options in solving the practical problem of para-
sitic reflection from the rear side of a transparent specimen [75]. Su
et al. proposed Software Configurable Optical Test System (SCOTS)
which has been used to measure large optics such as the primary
mirrors of telescopes [76, 77], and it demonstrated the first trial
to measure synchrotron mirrors with a full-field deflectometry sys-
tem [78–80]. In astronomy applications, Sironi et al. employed the
PMD technique to evaluate the free-form telescope mirrors [81, 82].
Häusler et al. discussed PMD technique from the physicist’s point of
view and the information theoretical point of view and compared it
with interferometry [83, 84]. Faber et al. gave a comparative assess-
ment of the strengths and weaknesses of PMD comparing to interfer-
ometry [85]. Liu et al. proposed Direct PMD (DPMD) which utilized
the geometric relations of parallel planes to directly calculate the
height from phase information and this approach can measure dis-
continuous specular objects [42, 86].
(2) Fast and dynamic measurements. Instead of using fringe orienta-
tion in one direction, Huang et al. applied two-dimensional crossed
fringe patterns in fringe reflection technique to measure dynamic de-
formation on specular surfaces [35]. With the advanced fringe pro-
cessing algorithms, both x- and y-slopes can be determined from a
single image acquisition [36]. Phase shifting algorithm for the two-
dimensional crossed fringe images was developed by Liu et al. for
applications allowing only a one-dimensional physical translation of
the crossed fringe pattern [44]. Liang et al. applied the single-shot
PMD technique to measure the corneal topography [87].
(3) Deformation measurement. Because the PMD technique is ex-
tremely sensitive to the out-of-plane deformations and its tolerance
to system calibration error is higher when measuring relative de-
formation, it becomes more and more attractive to apply the PMD
technique for deformation or defect measurement [36, 88, 89]. For
instance, the PMD technique has been used to measure influence
functions of a deformable mirror to perform closed-loop feedback
and control [90].
(4) Study on misalignment. Similarly, the influence of the misalign-
ment can also be determined by using PMD technique [91]. Davies
et al. utilized the PMD technique to determine the influences of 5°-
of-freedom misalignment of segmented telescope mirrors [92].
(5) Working beyond the visible light. With the success of PMD in the
visible region, it has been extended to other ranges of the wavelength
spectrum for specific applications. Sprenger et al. proposed a novel
method using Ultraviolet (UV) source in deflectometry to avoid the
parasitic reflections from the rear side of transparent specimens [93].
Su et al. explored deflectometry with Infrared (IR) source and camera
to measure rough optical surfaces [94]. Due to the lack of convenient
devices like the TFT LCD monitor in visible light to generate the
phase-shifted fringe patterns, instead of measuring the fringe phases,
the methods working with UV and IR sources calculate the intensity
peak while scanning a slit across the radiation source.
4. Related measurement techniques
There are several related techniques linked to the PMD technique.
Their similarities and differences are addressed in this work to view
these similar techniques from different angles.
4.1. Comparison with phase measuring profilometry
From the data acquisition to the image processing, the PMD is very
similar to the Phase Measuring Profilometry (PMP) [9, 18] which mea-
sures the 3D shape of diffused surfaces by using fringe projection. PMP
has a longer study history going back to early 1980 s [18]. In both
techniques, their raw data are commonly sinusoidal fringe patterns.
Their fringe analysis procedure similarly includes fringe demodulation
[19, 43] and phase unwrapping [46, 47, 50]. The fringe pattern(s) can
be demodulated by using least-squares-based phase shifting algorithms
[19] or single-frame demodulation techniques such as transform-based
methods [43]. Therefore, they share the same practical issues as well as
the existing solutions in fringe analysis, such as the nonlinear response
of the digital light device [95–98].
Although these two techniques measure the 3D shape based on the
geometry relation of the source and detectors. There is a major differ-
ence: PMP measures the diffuse surface based on the optical triangula-
tion as illustrated in Fig. 9(b), while PMD works with the specular re-
flection based on the law of reflection in Fig. 9(a). The retrieved phase
values in PMP are directly related to height data, while the phase values
in PMD are linked to both slopes and height. Therefore, proper regular-
ization and numerical integration are needed in PMD. When measuring
partially specular and diffuse surfaces, these two techniques can be com-
bined by using the PMP height data to regularize the inverse problem
of PMD [99, 100].
4.2. Comparison with pointwise scanning deflectometry
Based on the law of reflection, the PMD technique measures the full-
field SUT slopes from each measurement as shown in Fig. 10(a). Relying
on the same principle, the single-point deflectometry technique, such as
autocollimator illustrated in Fig. 10(b), only provides readings of the
intensity-weighted average x- and y-slopes in a small area in each mea-
surement.
The single-point deflectometry technique is commonly more pre-
cise than the full-field PMD technique. For instance, the measuring
precision is typically around 0.1 μrad RMS for the Long Trace Pro-
filer [101], Nanometer Optical component measuring Machine [102],
or Nano-accuracy Surface Profiler [103, 104] which are widely used
for the synchrotron mirror inspection, while the measuring precision of
full-field PMD technique is typically larger than 10 μrad RMS. Compar-
ing to the pointwise scanning deflectometry, the PMD technique is good
for full-field measurements with higher speed. Moreover, the field of
view and dynamic range of PMD systems are much easier to adjust for
different applications.
4.3. Comparison with Hartmann wavefront sensing
Turning to the system configuration and measuring process, the PMD
is similar to the Hartmann Wavefront Sensor (HWS) [105] in some as-
pects. As Su et al. [76] claimed, their SCOTS can be considered as a
reverse Hartmann test. The similarities between PMD and HWS include
that the incoming rays in HWS are similar to the camera probe rays in
PMD. By employing the fringe phase values, the screen in PMD is func-
tional as the detector in HWS as shown in Fig. 11. In both techniques,
the direct readouts in a single measurement are x- and y- slopes in a 2D
252
7. L. Huang et al. Optics and Lasers in Engineering 107 (2018) 247–257
Fig. 9. PMD (a) and PMP (b) share many similar aspects in pattern generation, image acquisition, and fringe analysis.
Fig. 10. PMD measures the full-field slopes (a) while the pointwise deflectometry measures the average slopes of a small area (b).
Fig. 11. The sampling point on the PMD can vary by the specimen distance (a), while the sampling points of HWS with respect to its detector is always fixed (b).
253
8. L. Huang et al. Optics and Lasers in Engineering 107 (2018) 247–257
grid with coordinates (x and y), or in a simple phrase, two pieces of slope
maps. The height or wavefront is calculated from x, y, x-slope, and y-
slope. Therefore, they share some technical aspects, such as wavefront
reconstruction algorithms. In fact, at the very beginning of the PMD
study, the 2D integration techniques are learned from the earlier wave-
front reconstruction work for the HWS [106]. Moreover, looking at the
potential post-processing on the measurement results, the surface shape
in PMD and the wavefront in HWS are commonly decomposed by us-
ing Zernike, Chebyshev, or Legendre polynomials for optical aberration
analysis.
However, their differences are obvious as well. The calibration in
HWS determines the geometric relations between the detector and the
pinholes on the Hartmann mask, while the calibration in PMD deter-
mines the relative geometric relations between the camera probe rays
and the screen. The sub-aperture pinholes in HWS are fixed once the
mask is mounted with the detector, so the sampling positions will not
change with respect to the detector. As illustrated in Fig. 11, although
the probe rays from the camera are consistent in PMD, the sampling
positions still can vary along the height level of the surface under test
(SUT). This difference will introduce two major issues in PMD. First, the
slope calculation in PMD is not as straightforward as it is in HWS which
always calculates slopes with the calibrated pinhole-detector distance.
Instead, by only knowing the geometric relation between camera probe
rays and display screen in PMD, the height-slope ambiguity exists when
calculating the slope with unfixed sampling position. It requires addi-
tional reasonable hypothesis or other means to regularize this ill-posed
problem. Second, the pinhole array of HWS is commonly in a regular
geometry, which makes the wavefront reconstruction relatively easier.
Although the 2D integration techniques in PMD originate from wave-
front reconstruction in HWS, the slope data in PMD have its own fea-
tures: owing to the unfixed sampling position and off-axis perspective,
the slope samplings in world coordinates are not in a regular grid, such
as the rectangle, but more generally in a quadrilateral grid. This issue
requires additional modifications on the classical wavefront reconstruc-
tion algorithm to reduce the integration error due to the irregular grids
[62].
5. Major challenges, current solutions, and remaining
deficiencies
As mentioned above, there are many technical challenges in PMD.
After years of research and development, some of the earlier difficulties
have been overcome with the current solutions, and some deficiencies
are still remaining and require more robust and flexible solutions.
5.1. Calibration
System calibration is and will always be a mandatory and critical
procedure before any metrology instrument able to deliver a reliable
measurement. The calibration process of a PMD can usually be divided
into the geometry calibration and the screen calibration. The purpose
of geometry calibration is to determine the geometric relations of the
display screen with respect to the camera rays (based on camera cali-
bration). Reference flat mirror with markers can be used to complete
the geometry calibration [23]. Inspired by the idea of bundle adjust-
ment as an iterative self-consistency approach, Olesch et al. developed
a self-calibration procedure for arbitrary specular surfaces to improve
the global accuracy of the reconstructed height data [34]. Xiao et al.
proposed a simple and flexible approach to carry out the PMD calibra-
tion with a marker-less flat mirror and optimization with bundle ad-
justment [107]. This method does not need any markers on the cali-
bration flat mirror, which is a great advantage because the accuracy
of the marker positions directly influences the calibration result in the
traditional method. Based on Xiao’s calibration method for monoscopic
PMD, Ren et al. further developed a calibration method for the stereo-
scopic PMD [108]. The stereoscopic PMD setup is first treated as two
separate monoscopic PMD systems and calibrated one by one with it-
erative optimization, and then they are merged into one cost function
to have an overall optimization. Instead of utilizing the PMD devices
only, laser tracker with retro-reflectors has been used to calibrate the
distances to assist the PMD calibration [79]. By using the laser tracker,
the calibration procedure gets more possibilities to define different ref-
erence planes and get distance and dimensions more flexible and easier,
instead of relying on camera vision only. Zhou et al. proposed a flexible
and simple PMD calibration method based on the combination of re-
flection rays determined by the varied points on a screen and reflection
images of a plane mirror without fiducials placed at three different loca-
tions [109]. In some strategies with additional geometry assumptions,
such as the parallel planes in DPMD [42, 86], particular alignment and
calibration techniques are needed to reduce the systematic error. How-
ever, since the low-frequency slope errors are much easier to propa-
gate into the height data as low-frequency profile errors via integration,
even with the existing PMD calibration approaches, there is still an open
question on how to easily determine the geometry relations with better
accuracy to fully utilize the high sensitivity of the slope measurement.
Nevertheless, it is noteworthy that because the operation of taking
the difference from the reference makes the deformation result less sen-
sitive to the imperfection of the calibration, the calibration requirement
for deformation measurements is lower than the calibration requirement
for absolute shape measurements [88]. This makes the PMD very suit-
able for the out-of-plane deformation measurement.
Screen imperfection was considered and a calibration procedure was
proposed in Refs. [23, 110]. In many PMD measurements reported in
the literature, the screen imperfection has not yet been well addressed
or carefully calibrated. With the further development in geometric cal-
ibration for PMD, at a certain point, the screen imperfection will need
to be carefully considered, or the PMD system may be more smartly set
up to minimize the influence by the screen imperfection.
5.2. Height-slope ambiguity
As mentioned earlier when introducing the setups of PMD in
Section 2.1, there are several approaches proposed to regularize the in-
verse problem, e. g. shape assumption [35–37], translated screens [26,
38, 39], distance regularization [41], and stereovision [24]. Besides,
The model-based optimization [111–114] is another way to deal with
this issue. In modal PMD (MPMD) [111], models with analytical expres-
sions are used to represent the shape of SUT. In this way, both height
and slopes are represented with a set of coefficients of modes, and the
ray tracing correspondence on the screen can be adjusted by changing
these coefficients. By optimizing these coefficients to minimize the dis-
crepancy between ray tracing and actual measurement, the slope and
height can be determined at the same time to best explain the measure-
ment data.
5.3. Rotational vectors
Theoretically, the gradient data of the SUT must be a conservative
vector field. However, due to the phase errors, and mainly the calibra-
tion errors and improper assumptions, the measured slope results con-
tain rotational vectors, instead of being a pure conservative vector field.
These rotational vectors will introduce problems to the following 2D in-
tegration process [115]. How to reduce this type of error is a challenging
task. Recently there is a method proposed by Xiao et al. for processing
the rotational vectors in deflectometry with sparse representation [116].
The MPMD can involve the screen pose to the optimization to extremely
reduce the residual rotational vectors [111]. Better initial value from
pre-calibration is always helpful to avoid a nonlinear optimization in
MPMD being stuck in a local minimum.
254
9. L. Huang et al. Optics and Lasers in Engineering 107 (2018) 247–257
5.4. Shape reconstruction from slopes (2D integration)
By learning from the wavefront reconstruction in wavefront sensing
[106] and after several years of investigation especially for PMD [57–
62, 70, 71, 73, 115, 117], the 2D integration for shape reconstruction
from slopes has been well developed.
The performance of the modal reconstruction in terms of accuracy
and speed mainly depends on how close the measured x-and y-slopes
can be fitted with a finite number of the selected modes in the chosen
mathematical model. The shape reconstruction becomes a representa-
tion by the selected modes weighted by the coefficient determined from
slope fitting process.
On the other hand, the performance of the zonal reconstruction is
mainly based on the selected format of the finite difference and the
sampling grid geometry of the slope data. A recent work on the zonal
reconstruction provides a general solution in a more common grid in
PMD with good performance in both accuracy and speed [62].
Many practical issues have been considered in the zonal and modal
reconstruction methods, e.g. the incomplete dataset with arbitrary aper-
ture [70], high accuracy [58, 61, 66], speed comparison [62, 115], and
irregular grids [62, 71]. A lot of reconstruction algorithms are ready to
be used in order to meet various requirements in accuracy and speed for
different applications.
6. Conclusions
With many years of research and development, a few aspects of the
PMD technique have been studied from its physical limitations and sys-
tem calibration to the regularization and shape reconstruction. As a high
dynamic range and low-cost measurement solution, the PMD technique
has been applied to several scientific applications for deformation, cur-
vature, and shape measurement. Accurate calibration for absolute shape
measurement still needs to be improved and it is a current limitation of
the PMD technique. Further investigations are still necessary to make
the PMD technique more flexible to calibrate, easier to use, and more
accurate in practical measurements for industrial applications.
Acknowledgment
This research used resources of the National Synchrotron Light
Source II, a U.S. Department of Energy (DOE) Office of Science User
Facility operated for the DOE Office of Science by Brookhaven National
Laboratory under Contract No. DE-SC0012704.
References
[1] Carbone V, Carocci M, Savio E, Sansoni G, De Chiffre L. Combination of a vision
system and a coordinate measuring machine for the reverse engineering of freeform
surfaces. Int J Adv Manuf Technol 2001;17:263–71.
[2] Leopold J, Günther H, Leopold R. New developments in fast 3D-surface quality
control. Measurement 2003;33:179–87.
[3] Wolf K, Roller D, Schäfer D. An approach to computer-aided quality control based
on 3D coordinate metrology. J Mater Process Technol 2000;107:96–110.
[4] Chu C-H, Song M-C, Luo VCS. Computer aided parametric design for 3D tire mold
production. Comput Ind 2006;57:11–25.
[5] Thiess H, Lasser H, Siewert F. Fabrication of X-ray mirrors for synchrotron appli-
cations. Nucl Instrum Methods Phys Res 2010;616:157–61 Section A: Accelerators,
Spectrometers, Detectors and Associated Equipment.
[6] Levoy M, Pulli K, Curless B, Rusinkiewicz S, Koller D, Pereira L, Ginzton M, An-
derson S, Davis J, Ginsberg J, Shade J, Fulk D. The digital Michelangelo project:
3D scanning of large statues. In: Proceedings of the twenty seventh annual confer-
ence on computer graphics and interactive techniques. ACM Press/Addison-Wesley
Publishing Co; 2000. p. 131–44.
[7] Harding K. Engineering precision. Nat Photonics 2008;2:667.
[8] Chen F, Brown GM, Song M. Overview of three-dimensional shape measurement
using optical methods. OPTICE 2000;39:10–22.
[9] Geng J. Structured-light 3D surface imaging: a tutorial. Adv Opt Photonics
2011;3:128–60.
[10] Chan VH, Bradley CH, Vickers GW. Automating laser scanning of 3D surfaces for
reverse engineering. Intell Syst Adv Manuf 1997:9 SPIE.
[11] Blais F. Review of 20 years of range sensor development. SPIE; 2004. p. 13.
[12] Cui Y, Schuon S, Chan D, Thrun S, Theobalt C. 3D shape scanning with a
time-of-flight camera. In: Proceedings of the IEEE computer society conference on
computer vision and pattern recognition; 2010. p. 1173–80.
[13] Werling S, Mai M, Heizmann M, Beyerer J. Inspection of specular and partially
specular surfaces. Metrol Meas Syst 2009;16:415–31.
[14] Franca JGDM, Gazziro MA, Ide AN, Saito JH. A 3D scanning system based on laser
triangulation and variable field of view. In: Proceedings of the IEEE international
conference on image processing; 2005 I-425-428.
[15] Xi F, Shu C. CAD-based path planning for 3-D line laser scanning. Comput Aided
Des 1999;31:473–9.
[16] Gorthi SS, Rastogi P. Fringe projection techniques: Whither we are? Opt Lasers Eng
2010;48:133–40.
[17] Burke J, Bothe T, Osten W, Hess CF. Reverse engineering by fringe projection. In:
Proceedings of the international symposium on optical science and technology.
SPIE; 2002. p. 13.
[18] Srinivasan V, Liu HC, Halioua M. Automated phase-measuring profilometry of 3-D
diffuse objects. Appl Opt 1984;23:3105–8.
[19] Huang PS, Zhang S. Fast three-step phase-shifting algorithm. Appl Opt
2006;45:5086–91.
[20] Surrel Y. Design of algorithms for phase measurements by the use of phase stepping.
Appl Opt 1996;35:51–60.
[21] Kemao Q. Windowed Fourier transform for fringe pattern analysis. Appl Opt
2004;43:2695–702.
[22] Kemao Q. Two-dimensional windowed Fourier transform for fringe pattern anal-
ysis: principles, applications and implementations. Opt Lasers Eng 2007;45:
304–317.
[23] Petz M, Tutsch R. Reflection grating photogrammetry: a technique for absolute
shape measurement of specular free-form surfaces, Optics and Photonics 2005,
SPIE2005, pp. 58691D.
[24] Knauer MC, Kaminski J, Häusler G. Phase measuring deflectometry: a new ap-
proach to measure specular free-form surfaces, Photonics Europe, SPIE2004, pp.
366-376.
[25] Bothe T, Li W, Kopylow Cv, Jüptner WPO. High-resolution 3D shape measurement
on specular surfaces by fringe reflection, Photonics Europe, SPIE2004, pp. 411-422.
[26] Petz M, Tutsch R. Measurement of optically effective surfaces by imaging of grat-
ings. In: Proceedings of SPIE; 2003. p. 288–94.
[27] Petz M, Ritter R. Reflection grating method for 3D measurement of reflecting sur-
faces. In: Proceedings of SPIE; 2001. p. 35–41.
[28] Balzer J, Werling S. Principles of Shape from Specular Reflection. Measurement
2010;43:1305–17.
[29] Tutsch R, Petz M, Keck C. Optical 3D measurement of scattering and specular re-
flecting surfaces. J Phys Conf Ser 2008;139:012008.
[30] Zhang Z, Wang Y, Huang S, Liu Y, Chang C, Gao F, Jiang X. Three-Dimensional
Shape Measurements of Specular Objects Using Phase-Measuring Deflectometry.
Sensors 2017;17:2835.
[31] Huang L, Wong JX, Asundi A. Specular 3D shape measurement with a compact
fringe reflection system. In: Proceedings of the international conference on optics
in precision engineering and nanotechnology (icOPEN2013). SPIE; 2013. p. 9.
[32] Maldonado AV, Su P, Burge JH. Development of a portable deflectometry system
for high spatial resolution surface measurements. Appl Opt 2014;53:4023–32.
[33] Li M, Li D, Jin C, E K, Yuan X, Xiong Z, Wang Q. Improved zonal integration method
for high accurate surface reconstruction in quantitative deflectometry. Appl Opt
2017;56:F144–51.
[34] Olesch E, Faber C, Häusler G. Deflectometric self-calibration for arbitrary specular
surfaces. In: Proceedings of the one hundred twelfth annual meeting of the DGaO
A; 2011.
[35] Huang L, Ng CS, Asundi AK. Dynamic three-dimensional sensing for specular sur-
face with monoscopic fringe reflectometry. Opt Express 2011;19:12809–14.
[36] Huang L, Seng Ng C, Asundi AKrishna. Fast full-field out-of-plane deformation mea-
surement using fringe reflectometry. Opt Lasers Eng 2012;50:529–33.
[37] Tang Y, Su X, Liu Y, Jing H. 3D shape measurement of the aspheric mirror by
advanced phase measuring deflectometry. Opt Express 2008;16:15090–6.
[38] Guo H, Tao T. Specular surface measurement by using a moving diffusive structured
light source. Photonics Asia 2007 2007:7 SPIE.
[39] Guo H, Feng P, Tao T. Specular surface measurement by using least squares light
tracking technique. Opt Lasers Eng 2010;48:166–71.
[40] Xiao Y-L, Li S, Zhang Q, Zhong J, Su X, You Z. Optical fringe-reflection deflectom-
etry with bundle adjustment. Opt Lasers Eng 2018;105:132–40.
[41] Li W, Sandner M, Gesierich A, Burke J. Absolute optical surface measurement with
deflectometry, SPIE Optical Engineering + Applications, SPIE2012, pp. 84940G.
[42] Liu Y, Huang S, Zhang Z, Gao N, Gao F, Jiang X. Full-field 3D shape measurement
of discontinuous specular objects by direct phase measuring deflectometry. Sci Rep
2017;7:10293.
[43] Huang L, Kemao Q, Pan B, Asundi AK. Comparison of Fourier transform, windowed
Fourier transform, and wavelet transform methods for phase extraction from a sin-
gle fringe pattern in fringe projection profilometry. Opt Lasers Eng 2010;48:141–8.
[44] Liu Y, Olesch E, Yang Z, Häusler G. Fast and accurate deflectometry with crossed
fringes. Adv Opt Technol 2014:441.
[45] Ghiglia DC, Pritt MD. Two-dimensional phase unwrapping: theory, algorithms, and
software. New York: Wiley; 1998.
[46] Su X, Chen W. Reliability-guided phase unwrapping algorithm: a review. Opt Lasers
Eng 2004;42:245–61.
255
10. L. Huang et al. Optics and Lasers in Engineering 107 (2018) 247–257
[47] Zhao M, Huang L, Zhang Q, Su X, Asundi A, Kemao Q. Quality-guided phase un-
wrapping technique: comparison of quality maps and guiding strategies. Appl Opt
2011;50:6214–24.
[48] Huntley JM, Saldner H. Temporal phase-unwrapping algorithm for automated in-
terferogram analysis. Appl Opt 1993;32:3047–52.
[49] Tian J, Peng X, Zhao X. A generalized temporal phase unwrapping algorithm for
three-dimensional profilometry. Opt Lasers Eng 2008;46:336–42.
[50] Zuo C, Huang L, Zhang M, Chen Q, Asundi A. Temporal phase unwrapping algo-
rithms for fringe projection profilometry: a comparative review. Opt Lasers Eng
2016;85:84–103.
[51] Zhang Z. A flexible new technique for camera calibration. IEEE Trans Pattern Anal
Mach Intell 2000;22:1330–4.
[52] Tsai R. A versatile camera calibration technique for high-accuracy 3D machine
vision metrology using off-the-shelf TV cameras and lenses. IEEE J Robot Autom
1987;3:323–44.
[53] Fried DL. Least-square fitting a wave-front distortion estimate to an array of
phase-difference measurements. J Opt Soc Am 1977;67:370–5.
[54] Hudgin RH. Wave-front reconstruction for compensated imaging. J Opt Soc Am
1977;67:375–8.
[55] Hudgin RH. Optimal wave-front estimation. J Opt Soc Am 1977;67:378–82.
[56] Southwell WH. Wave-front estimation from wave-front slope measurements. J Opt
Soc Am 1980;70:998–1006.
[57] Huang L, Asundi A. Improvement of least-squares integration method with iterative
compensations in fringe reflectometry. Appl Opt 2012;51:7459–65.
[58] Li G, Li Y, Liu K, Ma X, Wang H. Improving wavefront reconstruction accuracy by
using integration equations with higher-order truncation errors in the Southwell
geometry. J Opt Soc Am A 2013;30:1448–59.
[59] Ren H, Gao F, Jiang X. Improvement of high-order least-squares integration method
for stereo deflectometry. Appl Opt 2015;54:10249–55.
[60] Ren H, Gao F, Jiang X. Least-squares method for data reconstruction from gradient
data in deflectometry. Appl Opt 2016;55:6052–9.
[61] Huang L, Xue J, Gao B, Zuo C, Idir M. Spline based least squares integra-
tion for two-dimensional shape or wavefront reconstruction. Opt Lasers Eng
2017;91:221–6.
[62] Huang L, Xue J, Gao B, Zuo C, Idir M. Zonal wavefront reconstruction in
quadrilateral geometry for phase measuring deflectometry. Appl Opt 2017;56:
5139–5144.
[63] Dai G-m. Modal wave-front reconstruction with Zernike polynomials and
Karhunen–Loève functions. J Opt Soc Am A 1996;13:1218–25.
[64] Ares M, Royo S. Comparison of cubic B-spline and Zernike-fitting techniques in
complex wavefront reconstruction. Appl Opt 2006;45:6954–64.
[65] Dai F, Tang F, Wang X, Sasaki O, Feng P. Modal wavefront reconstruction based on
Zernike polynomials for lateral shearing interferometry: comparisons of existing
algorithms. Appl Opt 2012;51:5028–37.
[66] Mochi I, Goldberg KA. Modal wavefront reconstruction from its gradient. Appl Opt
2015;54:3780–5.
[67] Kewei E, Zhang C, Li M, Xiong Z, Li D. Wavefront reconstruction algorithm based
on Legendre polynomials for radial shearing interferometry over a square area and
error analysis. Opt Express 2015;23:20267–79.
[68] Ettl S, Kaminski J, Olesch E, Strauß H, Häusler G. Fast and robust 3D shape recon-
struction from gradient data. DGaO Proc 2007.
[69] Freischlad KR, Koliopoulos CL. Modal estimation of a wave front from dif-
ference measurements using the discrete Fourier transform. J Opt Soc Am A
1986;3:1852–61.
[70] Huang L, Idir M, Zuo C, Kaznatcheev K, Zhou L, Asundi A. Shape reconstruction
from gradient data in an arbitrarily-shaped aperture by iterative discrete cosine
transforms in Southwell configuration. Opt Lasers Eng 2015;67:176–81.
[71] Ettl S, Kaminski J, Knauer MC, Häusler G. Shape reconstruction from gradient data.
Appl Opt 2008;47:2091–7.
[72] Lowitzsch S, Kaminski J, Häusler G. Shape reconstruction of 3D objects from noisy
slope data, Proc. DGaO2005, pp. A22.
[73] Huang L, Asundi AK. Framework for gradient integration by combining radial basis
functions method and least-squares method. Appl Opt 2013;52:6016–21.
[74] Häusler G, Richter C, Leitz K-H, Knauer MC. Microdeflectometry—a novel tool to
acquire three-dimensional microtopography with nanometer height resolution. Opt
Lett 2008;33:396–8.
[75] Faber C, Knauer MC, Häusler G. Can deflectometry work in presence of parasitic
reflections?, Proc. DGaO2009, pp. A10.
[76] Su P, Parks RE, Wang L, Angel RP, Burge JH. Software configurable optical test
system: a computerized reverse Hartmann test. Appl Opt 2010;49:4404–12.
[77] Huang R, Su P, Horne T, Zappellini GB, Burge JH. Measurement of a large de-
formable aspherical mirror using SCOTS (software configurable optical test sys-
tem). SPIE Optical Engineering+ Applications, International Society for Optics and
Photonics; 2013. 883807-883807-883807.
[78] Su P, Wang Y, Burge JH, Kaznatcheev K, Idir M. Non-null full field X-ray mir-
ror metrology using SCOTS: a reflection deflectometry approach. Opt Express
2012;20:12393–406.
[79] Huang R, Su P, Burge JH, Huang L, Idir M. High-accuracy aspheric x-ray mirror
metrology using Software Configurable Optical Test System/deflectometry. OP-
TICE 2015;54 084103-084103.
[80] Su P, Khreishi M, Su T, Huang R, Dominguez MZ, Maldonado AV, Butel GP, Wang Y,
Parks RE, Burge JH. Aspheric and freeform surfaces metrology with software con-
figurable optical test system: a computerized reverse hartmann test. SPIE; 2013.
p. 11.
[81] Sironi G, Canestrari R, Pareschi G, Pelliciari C. Deflectometry for optics evaluation:
free form segments of polynomial mirror. In: Advances in optical and mechanical
technologies for telescopes and instrumentation. International Society for Optics
and Photonics; 2014. p. 91510T.
[82] Sironi G, Canestrari R, Tayabaly K, Pareschi G. Evaluation of novel approach to
deflectometry for high accuracy optics. Advances in optical and mechanical tech-
nologies for telescopes and instrumentation iI. International Society for Optics and
Photonics; 2016.
[83] Häusler G, Knauer MC, Faber C, Richter C, Peterhänsel S, Kranitzky C, Veit K.
Deflectometry challenges interferometry: 3-d-metrology from nanometer to meter,
advances in imaging, Vancouver: Optical Society of America; 2009. DMC4.
[84] Häusler G, Faber C, Olesch E, Ettl S. Deflectometry vs. interferometry. In: Proceed-
ings of the SPIE optical metrology. SPIE; 2013. p. 11.
[85] Faber C, Olesch E, Krobot R, Häusler G. Deflectometry challenges interferometry:
the competition gets tougher!. In: SPIE optical engineering + applications. SPIE;
2012. p. 15.
[86] Zhao P, Gao N, Zhang Z, Gao F, Jiang X. Performance analysis and evaluation of
direct phase measuring deflectometry. Opt Lasers Eng 2018;103:24–33.
[87] Liang H, Olesch E, Yang Z, Häusler G. Single-shot phase-measuring deflectometry
for cornea measurement. Adv Opt Technol 2016:433.
[88] Li W, Huke P, Burke J, von Kopylow C, Bergmann RB. Measuring deformations
with deflectometry, 2014, pp. 92030F-92030F-92012.
[89] Asundi A, Lei H, Eden TKM, Sreemathy P, May WS. Phase shift reflectometry for
sub-surface defect detection. Photonics Asia, SPIE; 2012. p. 4.
[90] Huang L, Zhou C, Zhao W, Choi H, Graves L, Kim D. Close-loop performance of a
high precision deflectometry controlled deformable mirror (DCDM) unit for wave-
front correction in adaptive optics system. Opt Commun 2017;393:83–8.
[91] Zhao W, Huang R, Su P, Burge JH. Aligning and testing non-null optical system
with deflectometry, SPIE Optical Engineering+ Applications, International Society
for Optics and Photonics 2014, pp. 91950F-91950F-91959.
[92] Davies A, Vann T, Evans C, Butkiewicz M. Phase measuring deflectometry for de-
termining 5 DOF misalignment of segmented mirrors, Applied Optical Metrology
II, International Society for Optics and Photonics 2017, pp. 103730H.
[93] Sprenger D, Faber C, Seraphim M, Häusler G. UV-Deflectometry: No parasitic re-
flections. In: Proc. DGaO; 2010. p. A19.
[94] Su T, Wang S, Parks RE, Su P, Burge JH. Measuring rough optical surfaces using
scanning long-wave optical test system. 1. Principle and implementation. Appl Opt
2013;52:7117–26.
[95] Guo H, He H, Chen M. Gamma correction for digital fringe projection profilometry.
Appl Opt 2004;43:2906–14.
[96] Zhang S, Yau S-T. Generic nonsinusoidal phase error correction for three-dimen-
sional shape measurement using a digital video projector. Appl Opt 2007;46:36–43.
[97] Pan B, Kemao Q, Huang L, Asundi A. Phase error analysis and compensation for
nonsinusoidal waveforms in phase-shifting digital fringe projection profilometry.
Opt Lett 2009;34:416–18.
[98] Hoang T, Pan B, Nguyen D, Wang Z. Generic gamma correction for accuracy en-
hancement in fringe-projection profilometry. Opt Lett 2010;35:1992–4.
[99] Huang L, Asundi A. Study on three-dimensional shape measurement of partially
diffuse and specular reflective surfaces with fringe projection technique and fringe
reflection technique. In: SPIE optical engineering + applications. SPIE; 2011. p. 7.
[100] Sandner M. Hybrid Reflectometry–3D shape measurement on scattering and reflec-
tive surfaces. In: Proceedings of the DGaO; 2014.
[101] Takacs PZ, Qian S-n, Colbert J. Design of a long trace surface profiler. In: Pro-
ceedings of the OE LASE’87 and EO imaging symposium. International Society for
Optics and Photonics; 1987. p. 59–64. (January 1987, Los Angeles).
[102] Siewert F, Noll T, Schlegel T, Zeschke T, Lammert H. The nanometer optical com-
ponent measuring machine: a new sub‐nm topography measuring device for x‐ray
optics at BESSY. AIP Conf Proc 2004;705:847–50.
[103] Qian S, Gao B. Nano-accuracy measurement technology of optical-surface profiles,
2016, pp. 96870E-96870E-96810.
[104] Qian S, Idir M. Innovative nano-accuracy surface profiler for sub-50 nrad rms mir-
ror test, 2016, pp. 96870D-96870D-96810.
[105] Tarabocchia M, Holly S. Dynamic Hartmann wavefront sensor in applications. In:
Proceedings of the SPIE; 1982. p. 93–100.
[106] Li W, Bothe T, Kopylow Cv, Juptner WPO. Evaluation methods for gradient mea-
surement techniques. In: Proceedings of the photonics Europe, SPIE; 2004. p. 12.
[107] Xiao Y-L, Su X, Chen W. Flexible geometrical calibration for fringe-reflection 3D
measurement. Opt Lett 2012;37:620–2.
[108] Ren H, Gao F, Jiang X. Iterative optimization calibration method for stereo deflec-
tometry. Opt Express 2015;23:22060–8.
[109] Zhou T, Chen K, Wei H, Li Y. Improved system calibration for specular surface
measurement by using reflections from a plane mirror. Appl Opt 2016;55:7018–28.
[110] Maestro-Watson D, Izaguirre A, Arana-Arexolaleiba N. LCD screen calibration for
deflectometric systems considering a single layer refraction model. In: Proceedings
of the IEEE international workshop of electronics, control, measurement, signals
and their application to mechatronics (ECMSM); 2017. p. 1–6.
[111] Huang L, Xue J, Gao B, McPherson C, Beverage J, Idir M. Modal phase measuring
deflectometry. Opt Express 2016;24:24649–64.
[112] Huang L, Xue J, Gao B, McPherson C, Beverage J, Idir M. Model mismatch anal-
ysis and compensation for modal phase measuring deflectometry. Opt Express
2017;25:881–7.
[113] Liu M, Hartley R, Salzmann M. Mirror surface reconstruction from a single image.
In: Proceedings of the IEEE conference on computer vision and pattern recognition;
2013. p. 129–36.
256
11. L. Huang et al. Optics and Lasers in Engineering 107 (2018) 247–257
[114] Liu M, Hartley R, Salzmann M. Mirror surface reconstruction from a single image.
IEEE Trans Pattern Anal Mach Intel 2015;37:760–73.
[115] Huang L, Idir M, Zuo C, Kaznatcheev K, Zhou L, Asundi A. Comparison of two-di-
mensional integration methods for shape reconstruction from gradient data. Opt
Lasers Eng 2015;64:1–11.
[116] Xiao Y-L, Li S, Zhang Q, Zhong J, Su X, You Z. Optical fringe-reflection deflectom-
etry with sparse representation. Opt Lasers Eng 2018;104:62–70.
[117] Zhou T, Chen K, Wei H, Li Y. Improved method for rapid shape recovery
of large specular surfaces based on phase measuring deflectometry. Appl Opt
2016;55:2760–70.
257