See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/225179627

On image pre-processing for PIV of single- and two-phase flows over

reflecting objects

Article  in  Experiments in Fluids · August 2010

DOI: 10.1007/s00348-010-0827-y

CITATIONS READS

26 217

8 authors, including:

Niels G Deen Paul Willems

Eindhoven University of Technology University of Twente
229 PUBLICATIONS   6,664 CITATIONS    13 PUBLICATIONS   296 CITATIONS   

SEE PROFILE SEE PROFILE

M. Van Sint Annaland Hans A. M. Kuipers

Eindhoven University of Technology Eindhoven University of Technology
369 PUBLICATIONS   8,399 CITATIONS    577 PUBLICATIONS   17,708 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Desalination View project

ERC research project Multi-Scale Modeling of Mass and Heat Transfer in Dense Gas-Solid Flows View project

All content following this page was uploaded by Paul Willems on 16 May 2014.

The user has requested enhancement of the downloaded file.

Exp Fluids
DOI 10.1007/s00348-010-0827-y

RESEARCH ARTICLE

On image pre-processing for PIV of single- and two-phase flows

over reflecting objects
Niels G. Deen • Paul Willems • Martin van Sint Annaland •
J. A. M. Kuipers • Rob G. H. Lammertink • Antoine J. B. Kemperman •

Matthias Wessling • Walter G. J. van der Meer

Received: 19 August 2009 / Revised: 12 January 2010 / Accepted: 14 January 2010

Ó The Author(s) 2010. This article is published with open access at Springerlink.com

Abstract A novel image pre-processing scheme for PIV practice, however, these conditions cannot always be
of single- and two-phase flows over reflecting objects accomplished. Laser intensity can vary between images or
which does not require the use of additional hardware is image pairs due to differences in the two lasers (in case of a
discussed. The approach for single-phase flow consists of double-pulsed YAG laser), objects or bubbles can intro-
image normalization and intensity stretching followed by duce strong reflections of light, and/or reflection from
background subtraction. For two-phase flow, an additional channel walls in confined flows introduce glow. Several
masking step is added after the background subtraction. approaches are available in literature to tackle these
The effectiveness of the pre-processing scheme is shown problems (Seol and Socolofsky 2008; Lindken and Mer-
for two examples: PIV of single-phase flow in spacer-filled zkirch 2002; Honkanen and Nobach 2005; Westerweel
channels and two-phase flow in these channels. The pre- 1993; Shavit et al. 2007; Theunissen et al. 2008). In the
processing scheme increased the displacement peak presence of static objects or in two-phase flow, fluorescent
detectability significantly and produced high quality vector particles and a color filter are often used to avoid inter-
fields, without the use of additional hardware. ference of the object edges in the correlation map. In the
case of two-phase flow, two cameras (of which one has a
color filter) may be used to optically separate the two
1 Introduction phases before correlation. Although this enables PIV of
two-phase flow and edge flow, the average intensity of the
Throughout the years, digital particle imaging velocimetry tracer particles is lower (Raffel et al. 2007). In some cases,
(PIV) has developed into a technique that can be used in this will render the correlation rather difficult. Furthermore,
more and more complicated systems (Adrian 2005). It has fluorescent particles are about a factor ten more expensive
been successfully applied in turbulent flows, two-phase than non-fluorescent particles, and calibration of the two
flows, and flows around objects. The performance of the cameras is more complicated (Seol and Socolofsky 2008).
technique is determined by the quality of the images and by Applying a static mask to block stationary objects from
the signal treatment after acquisition (Raffel et al. 2007). If an image is a well-known technique to prevent these objects
the quality of the original images is good, i.e., even illu- from interfering with the correlation map. However, if an
mination, good contrast, low background noise, few sta- object is semi-transparent, information on flow behind the
tionary objects, suitable tracer particle displacement etc., object will be lost. To cope with moving bubbles shadow-
image processing will be relatively straightforward. In graphy is often used (Lindken and Merzkirch 2002). This
approach uses a second camera and background lighting to
N. G. Deen (&)
P. Willems
M. van Sint Annaland
capture the shadows of moving objects, which can then be
J. A. M. Kuipers
R. G. H. Lammertink
used to mask these areas from the images from the first
A. J. B. Kemperman
M. Wessling
W. G. J. van der Meer camera. However, this approach requires an additional
Faculty of Science and Technology, Institute of Mechanics,
camera and careful alignment of the images from the two
Processes and Control Twente (IMPACT), University of Twente,
Enschede, The Netherlands cameras. Since a large difference in size and intensity
e-mail: [email protected] between the object and tracer particles is often present,

123
 Exp Fluids

these objects can also be identified and masked using only where Imin(x) and Imax(x) are, respectively, the local
the original images from a single camera. sliding minimum and maximum using a filter length that
The use of an additional camera can be circumvented by is generally larger than the particle image diameter
removing the stationary objects through a background sub- (*3 px) and smaller than the interrogation window width
traction, which will leave moving tracer particles in the image (*32 px):
as suggested by Honkanen and Nobach (2005). Image nor- Imin ðxÞ ¼ minðIðxÞÞ; Imax ðxÞ ¼ maxðIðxÞÞ ð2Þ
malization to cope with uneven illumination was already x2X x2X

suggested by Westerweel (1993). This methodology scales where X is the square-shaped filter domain.
the intensity in the original image to a suitable minimum and/ To ensure that the local minima and maxima have
or maximum value. Common examples are subtracting a the same intensities, the filtered intensities are stretched
sliding minimum, sliding average, or scaling to the sliding based on the global minimum and maximum intensities:
minimum and -maximum. The last option is very attractive in
NðxÞ  Imin
applications with low light intensity, since the technique SðxÞ ¼ : ð3Þ
enhances tracer particle visibility. Since this approach also Imax  Imin
reduces the relative intensity of bright objects (bubble or After the image is properly normalized and stretched,
object reflections) compared to the particle intensities, the frames 1 and 2 can be subtracted from one another to
relative contribution of these objects in the correlation func- remove stationary objects in the background:
tion is also reduced (Shavit et al. 2007). Combining image
B1 ðxÞ ¼ maxðS1 ðxÞ  S2 ðxÞ; 0Þ;
normalization with background subtraction solves the prob- ð4Þ
B2 ðxÞ ¼ maxðS2 ðxÞ  S1 ðxÞ; 0Þ:
lems of temporal and spatial variation in the intensity distri-
bution, as discussed by Theunissen et al. (2008) recently.
In this work, we introduce a combined approach that can 3 Mask generating for images containing
be implemented at relative ease, which tackles uneven non-stationary objects
illumination, the presence of stationary objects and moving
objects (i.e. bubbles) without the use of additional hard- When non-stationary objects, like bubbles, are also present
ware. This approach consists of intensity normalization (to in the images, the algorithm for stationary objects requires
cope with uneven illumination), followed by background one additional step. While the normalization, stretching
subtraction (to remove stationary objects) and image and background subtraction is sufficient to remove the
masking (to remove the bubbles in two-phase flow). We stationary objects, the removal of non-stationary objects is
demonstrate the capabilities of this approach with two accomplished by applying a mask to the moving objects.
examples: single-phase flow in spacer-filled channels and This procedure was specifically developed for the removal
an extension to two-phase flow in these channels. of bubbles, which appear as bright rings on the image and
is similar to Seol and Socolofsky (2008). When other
moving objects are present, the generation of the mask
2 Theory might require adaptations depending on the reflection pat-
tern of the object (Fig. 1).
First the contrast of each frame is normalized with the aid To create the mask, a frame that does not contain any
of a local min/max filter: bubbles is subtracted from the frame to be analyzed.
IðxÞ  Imin ðxÞ Subsequently, the resulting image is scaled to (i.e. divided
NðxÞ ¼ ð1Þ by) the maximum intensity of the image, yielding an image
Imax ðxÞ  Imin ðxÞ

Single phase flow

I N S Background B ν
Normalization Stretching PIV
subtraction

Two phase flow

I N S Background B M ν
Normalization Stretching Masking PIV
subtraction

BW
Mask generation

Fig. 1 Schematic overview of the preprocessing steps

123
Exp Fluids

with intensities between 0 and 1. Then, the scaled image is contain openings, which are closed by applying an erode-
binarized, using a threshold that is determined by trial and dilate operation three times to ensure full closure of the
error giving a typical value of 0.02. That is, each pixel with bubble edge, using a circular structuring element of 20 px.
intensity higher than the threshold is set to 1. Subsequently, Now, all bubble edges should form closed rings, and the
noise is removed (i.e., small objects with less than 200 bubbles can be filled up by setting the intensity of the
connecting pixels). The obtained image now contains only bubble interior equal to 1. In a final step, any small objects
bright rings originating from the bubbles. These rings may are removed from the image, and the image is inverted so
that the bubbles have an intensity value of 0, and all the rest
have an intensity value of 1.

4 Results and discussion

4.1 Liquid flow

As a first example of the pre-processing method described

in this paper, single-phase liquid flow through a spacer-
filled channel is discussed. A spacer is a mat-like structure
with two layers of filaments at non-zero angles relative to
each other, see for example Fig. 2. Generally, angles vary
between 45 and 90°. These spacers are typically used to
separate membrane sheets and promote turbulence in a
membrane module. This type of structure is encountered in
spiral wound modules, which are often used for drinking
water production (Schwinge et al. 2004) (also see Fig. 3).
Figure 4a–d show the correlation map at the position
indicated in Fig. 2 for various stages of the pre-processing
sequence. This area includes parts of the spacer structure.
Fig. 2 Raw image, filaments \\ are at the front of the cell, filaments //
Figure 4a shows the correlation map for the original image.
are at the back. The square indicates the position of the interrogation This shows one broad peak with the center at the origin.
area shown in this article (located at 788, 988 px) This is mainly caused by the correlation of the stationary

Fig. 3 Schematic
representation of a spiral-wound
membrane module (taken from
www.mtrinc.com with
permission)

123
 Exp Fluids

Fig. 4 1289128 Correlation map for a the original image, b the -11 px)). Note that the original image is indicated by a white square
background subtracted image, c the normalized image, and d the in Fig. 2. The xy positions are indicated in pixels
normalized & background subtracted image (peak position (-12,

spacer, which is oriented at a 45° angle to the x-direction. in the correlation map, it is not detected by the PIV algo-
Figure 4b shows that after background subtraction, no clear rithm. The height of the correlation peak in general is also
displacement peak is present; however, a displacement limited due to the varying intensities between the two laser
peak is visible in a valley. Since this peak is not the highest pulses. The correlation map for the normalized image

Fig. 5 Time-averaged velocity field (final interrogation area

32932 px, 100 frames), the inverted original image is shown in the Fig. 6 Raw image for gas–liquid flow, the square indicates the
background, and only 1/4 of the vectors is shown for clarity interrogation area (position 547, 547 px)

123
Exp Fluids

Fig. 7 1289128 Correlation map for a the original image, b the subtracted, and masked image (peak position (4, -6 px)). Note that
masked image, c the normalized image, d the normalized & the original image is indicated by a white square in Fig. 6. The xy
background subtracted image, and e the normalized, background positions are indicated in pixels

(Fig. 4c) shows a large ridge at a 45° angle to the x- fouled by for example deposits and biofouling. Recently,
direction with a sharp peak at the origin, which is again the use of air sparging has been suggested as a cleaning
caused by the spacer. The displacement peak is only visible method for these modules (Cui et al. 2003). Several options
once the normalization and background subtraction are have been proposed for the mechanism of cleaning method
combined, as shown in Fig. 4d. The resulting time-aver- through air sparging, which mechanism dominates is still
aged vector field is shown in Fig. 5. The problem of unknown, and optimization of the cleaning process is
elimination of particles with a displacement smaller than mainly empirical. Therefore, a detailed study on two-phase
one particle diameter by the image subtraction algorithm flow in these channels is of interest to ultimately improve
(Theunissen et al. 2008) is not applicable here, since the drinking water capacity and quality.
magnification combined with the size of the applied The effect of normalization, subtraction, and masking
interrogation area is not detailed enough to resolve the for the position shown in Fig. 6 is shown in Fig. 7. The
boundary layer flow around the spacer. correlation map of the original image (Fig. 7a) shows a
large ridge, which corresponds with the bubble edge. Fig-
4.2 Gas–liquid flow ure 7b shows that masking does not remove this ridge, due
to the uneven illumination of the image pairs. If the image
Conventionally, spiral wound modules are operated in is only normalized (Fig. 7c), this ridge is still present,
single-phase flow. During operation, these modules get although the overall height of the peak is reduced. This is

123
 Exp Fluids

the presence of stationary objects by normalization fol-

lowed by background subtraction. This has the advantage
over masking of the stationary objects that flow behind/
over a semitransparent object can still be studied. For two-
phase flow, this procedure was extended with a masking
step to remove non-stationary bubbles from the image pair.
In both cases, all the steps in the procedure are required to
obtain the particle displacement peak. The velocity fields
that can be obtained for spacer-filled channels with this
procedure will provide valuable information to determine
the mechanism behind the cleaning through air sparging as
well as high quality validation data for CFD models. In the
long term, this information can be used to optimize air
sparging in industrial applications.

Acknowledgments The authors would like to acknowledge IMPACT

for financially supporting this work.

Open Access This article is distributed under the terms of the

Creative Commons Attribution Noncommercial License which per-
mits any noncommercial use, distribution, and reproduction in any
Fig. 8 Vector field of the image shown in Fig. 6 (final interrogation medium, provided the original author(s) and source are credited.
area 32932 px), the inverted original image is shown in the
background

due to the reduction in the bubble intensity relative to the References

particle and spacer intensity by the normalization proce-
dure. As can be seen from Fig. 7d, this problem is not Adrian RJ (2005) Twenty years of particle image velocimetry. Exp
solved by the subsequent background subtraction. Since the Fluids 39:159
Cui ZF, Chang S, Fane AG (2003) The use of gas bubbling to enhance
bubble is moving, parts of the bubble edge are still present membrane processes. J Membr Sci 221:1
in the subtracted image, and these interfere in the corre- Honkanen M, Nobach H (2005) Background extraction from double-
lation of the particles. Again, the displacement peak of the frame PIV images. Exp Fluids 38:348
particles is only detectable in the final step of the process, Lindken R, Merzkirch W (2002) A novel PIV technique for
measurements in multiphase flows and its application to two-
which is after the bubble masking. This removes the bubble phase bubbly flows. Exp Fluids 33:814
residues from the normalized & background subtracted Raffel M, Willert CE, Wereley ST, Komenhaus J (2007) Particle
image, where after only the particles remain in the image. inage velocimetry—a practical guide, 2nd edn. Springer, Berlin
This example shows that a combined approach can sig- Schwinge J, Neal PR, Wiley DE, Fletcher DF, Fane AG (2004) Spiral
wound modules and spacers: review and analysis. J Membr Sci
nificantly enhance the correlation of the images with 242:129
challenging circumstances. The vector field for this image Seol DG, Socolofsky SA (2008) Vector post-processing algorithm for
pair is shown in Fig. 8. phase discrimination of two-phase PIV. Exp Fluids 45:223
Shavit U, Lowe RJ, Steinbuck JV (2007) Intensity Capping: a simple
method to improve cross-correlation PIV results. Exp Fluids
42:225
5 Conclusions Theunissen R, Scarano F, Riethmuller M (2008) On improvement of
PIV image interrogation near stationary interfaces. Exp Fluids
It was shown that the processing of PIV-image pairs can be 45:557
Westerweel J (1993) Digital image particle velocimetry—theory and
greatly enhanced by using a proper pre-processing scheme. application. Delft University of Technology, Delft
For single-phase flow, good correlation can be obtained in

123

View publication stats