The document presents a digital image stabilization (DIS) technique based on the Hilbert-Huang transform (HHT). The technique has three main steps: (1) estimating local motion vectors (LMVs) in an image sequence, (2) using empirical mode decomposition (EMD) as part of HHT to decompose the LMV signals into intrinsic mode functions (IMFs), and (3) estimating jitter motion vectors from the IMFs to remove unwanted shaking from the image sequence. The technique is tested on three image sequences, with the LMV and IMF results presented. The HHT-based DIS method effectively separates camera intentional motion from unwanted jitter motion for stabilized video.
Report
Share
Report
Share
1 of 6
Download to read offline
More Related Content
Da4201691696
1. Sonali Dhanorkar et al Int. Journal of Engineering Research and Applications
ISSN : 2248-9622, Vol. 4, Issue 2( Version 1), February 2014, pp.691-696
RESEARCH ARTICLE
www.ijera.com
OPEN ACCESS
Determination of IMF using Hilbert–Huang Transform
Sonali Dhanorkar*, Dr.Y.S.Angal**
*,**(Department ofElectronics and telecommunication ,Pune University,India.)
ABSTRACT
In this paper, DIS technique is presented. DIS is proposed to stably remove the unwanted shaking phenomena in
the image sequences captured by cameras without the influence caused by moving object in the image or
intentional motion. Local Motion Vector estimation technique is used Sum Of absolute difference (SAD)
method . local motion vectors(LMV) of an image sequence are calculated . LMV of image sequence is used for
DIS Technique, which is based on the Hilbert–Huang transform (HHT) is proposed. The HHT technique contain
main block is empirical mode decomposition (EMD).The calculated Image sequence of an local motion vectors
are processed by the HHT in order to define both signals. The real Signal is divided into a number of
waveforms, called intrinsic mode functions (IMFs), using the process of empirical mode decomposition.
Keywords - DIS,EMD,HHT,Image sequence,IMFs, LMV.
I.
Introduction
Digital image sensor, such as handheld cameras,
mobile phones, and robots, are equipped with variety
of embedded systems which can produce image
sequences. The produced image sequences contains
motion caused by two different types of movements:
the smooth camera motion (intentional) and the
unwanted
shaking
motion(jitter).
Image stabilization is a form of technology used to
stabilize an image for a much clearer picture, which
can be include in a camera, important for producing
the sharpest photos. The image stabilization process
aims at removing irregular motion phenomena from
image sequences in order to accomplish a
compensated sequence that displays smooth camera
movements [1].The image stabilization systems can
be classified into three major types: the electronic,
the optical, and the digital stabilizers. The electronic
image stabilizer (EIS): Camera attached electronic
image stabilizer (EIS) uses Motion gyroscope sensor
for detecting camera movement. Optical image
stabilizers (OIS): Video cameras that stabilize the
recorded image by varying the optical path to the
sensor are used in Optical image stabilizers
(OIS).Because both EIS and OIS are hardware
dependent, the applications are restricted to device
built-in online processes. Digital image stabilization
(DIS):Digital image stabilization (DIS) is the process
of removing the undesired motion effects to generate
a compensated image sequence by using digital
image processing techniques without any mechanical
devices such as gyro sensors or a fluid prism [2].The
major advantages of DIS are: 1) machine
independence and 2) suitability for hardware
implementation [2].
www.ijera.com
II.
Local Motion Vector(LMV)
The motion estimation unit and the motion
compensation unit are two processing units of the
DIS system .To estimate the reliable global camera
movement through the acquired image sequence is
main purpose of motion estimation unit .To estimate
the local motion vectors (LMV) had been developed
various algorithms such as representative point
matching (RPM), edge pattern matching (EPM), bitplane matching (BPM) and others. The major
objective of these algorithms is to reduce the
computational complexity. LMVs are calculated
during the process of motion estimation within
smaller frame regions Essentially, LMVs represent
the offset of specific image regions between two
consecutive frames. Thus, LMVs include both the
intentional and the unwanted motion of the camera.
Figure.1.typical DIS system are divided into three
stages, in the first stage, LMV is estimated within a
specific frame region. The segregation of the
previous estimated LMV into the intentional and the
unwanted camera movement is dedicated in second
stage. After the segregation unit, the image sequence
is further processed in the image compensation unit
where the high-frequency movement is removed
from each frame[1].
This paper is organized as follows. In Section 3 ,HHT
includes EMD process is presented. In Sections 4, the
experimental results are provided, respectively.
Finally, conclusions are drawn in Section 5.
691 | P a g e
2. Sonali Dhanorkar et al Int. Journal of Engineering Research and Applications
ISSN : 2248-9622, Vol. 4, Issue 2( Version 1), February 2014, pp.691-696
Fig.1 typical DIS System after acquiring the image sequence.
III.
Hilbert Huang Transform(HHT)
Hilbert-Huang Transform (HHT) is a data analysis
tool, first developed in 1998, which is National
Aeronautics and Space Administration’s. It is
designated name for the combination of the EMD and
the HSA. Empirical mode decomposition(EMD)
divided into basis functions called intrinsic mode
functions (IMFs). Combination of the EMD and the
Hilbert spectral analysis (HSA) is HHT; the HSA
includes the Hilbert transform of each IMF generated
by the EMD process [1].
HHT have three main blocks:
3.1. EMD
3.2. Hilbert Transform
3.3.Jitter Motion Vector Estimation
www.ijera.com
Figure.3.shows Flowchart of EMD Process. The
sifting process is repeated until the signal meets the
definition of an IMF. Then, the IMF is subtracted
from the original signal, The process to generate one
IMF considered as an inner loop, as shown in
Figure.3. and the sifting process is repeated on the
remainder. The residue is treated as the new data and
subjected to the same sifting process (start of outer
loop in Figure.3.) This is repeated until the final
residue is a monotonic function. The last extracted
IMF is the lowest frequency component of the signal,
better known as the trend. Terminating criteria are
applied to the sifting process for IMFs since allowing
sifting to go beyond a certain point. The entire EMD
process is terminated if any of the following criteria
is satisfied:
when the residue is a function with one unique
extremum.
when the residue becomes a monotonic function
from which no IMF can be extracted [1].
The sum of the IMFs and the residue recovers the
original signal, which indicates completeness.
3.1 EMD
Fig.2.Shows DIS method using HHT
In general, The first component to the HHT method
is EMD algorithm .The EMD separates nonstationary
data into locally nonoverlapping time-scale
components. EMD attempts to decompose any signal
into a finite set of functions, whose Hilbert
transforms give physical instantaneous frequency
values. These functions are called intrinsic mode
functions (IMFs)[1].
EMD Algorithm :The sifting process is as follows:
1) Identified the local extrema (maxima, minima) of
the signal.
2) Connect the maxima with an interpolation
function, creating an upper envelope about the signal.
3) Connect the minima with an interpolation
function, creating a lower envelope about the signal.
4) Calculate the mean of the upper and lower
envelopes.
5) Subtract the local mean from the original signal.
Iterate on the residual
6) The residue is treated as the new data and
subjected to the same sifting process (start of outer
loop in Fig. 3)[4].
www.ijera.com
Fig..3.shows Flowchart of EMD Process
IV.
Experimental Result:
To verify the effectiveness of the proposed DIS
method, The experimental results of proposed system
are presented in this article. The horizontal
displacements and vertical displacements are
presented, the procedure for horizontal motions is
exactly the same.
In order to evaluate the performances of the method,
three different image sequences are processed. In this
paper, three image sequences are used to find out
IMF and horizontal ,vertical LMV. In Ist image
sequence is shaky car video which is available in
video library of Matlab. It has 10 sec duration, 132
692 | P a g e
3. Sonali Dhanorkar et al Int. Journal of Engineering Research and Applications
ISSN : 2248-9622, Vol. 4, Issue 2( Version 1), February 2014, pp.691-696
frames. Frame rate is 13.2frame/sec. the frame size
can be depends on duration of image sequence. In II nd
image sequence is walking man ,it is taken from web
camera 1.3Mpixel ,10sec duration and 132 pixel size.
In Ist image sequence is moving car. It has 216
frames,15 sec duration.The resulting vertical and
horizontal LMV from an image sequences is shown
in Fig 4(a),5(a),6(a) .the resulting vertical and
horizontal LMV calculate translational movement.
The upper and lower envelope are created using the
local maxima and local minima of initial signal. mean
envelope can be calculated from upper and lower
envelopes fig 4(b),5(b),6(b)Horizontal (X) LMV
Signal ,upper and lower envelopes and mean signal
and 4(d),5(d),6(d) vertical (Y) LMV Signal ,upper
and lower envelopes and mean signal.The resulting
IMF shown in fig.4(c),5(c),6(c) of X signal and
fig4(e)5(e)6(e) of Y signal.
V.
Conclusion:
DIS method based on the HHT has been presented.
LMV and EMD , which is first block of HHT has
been presented. By using Image sequences achieve
the definition of two fundamental motion such as
horizontal and vertical LMV.EMD results also
achieve such as upper and lower envelope, mean and
first IMF.
Table I:
Resulting values of DIS algorithms for all the tested
image sequences.
Image
Displacement
Displacement
Sequences (Pixel) at
(Pixel) at frame
Frame 60
131
X
Y X
Y
-20
28 -31
-64
Ist
-16
-46 -20
-26
IInd
-4
37 -64
64
IIIrd
www.ijera.com
VI.
www.ijera.com
Work To Be Implemented:
Implementing work is Hilbert transform ,which can
be used to generating various IMF. Observedmotion
caused by two different types of movements: the
smooth camera motion (intentional) and the
unwanted shaking motion (jitter).
REFERENCES
[1] Konstantinos Ioannidis and Ioannis Andreadis
“A Digital Image Stabilization Method Based
on the Hilbert–Huang Transform” IEEE
Transactions
On Instrumentation
And
Measurement, Vol. 61, No. 9, September 2012
[2] Sheng-Che Hsu, Sheng-Fu Liang, Kang-Wei
Fan, and Chin-Teng Lin, Fellow, Ieee” A
Robust In-Car digital Image Stabilization
Technique”IEEE Transactions On Systems,
Man, And Cybernetics—Part C: Applications
And Reviews, Vol. 37, No. 2, March 2007
[3] Sheng-Che Hsu, Sheng-Fu Liang, and ChinTeng Lin” A Robust Digital Image
Stabilization Technique Based on Inverse
Triangle Method and Background Detection”
IEEE Transactions on Consumer 336
Electronics, Vol. 51, No. 2, May 2005
[4] C. Caraffi, S. Cattani, and P. Grisleri, “Offroad path and obstacle detection using
decision networks and stereo vision,” IEEE
Trans. Intell.Transp. Syst., vol. 8, no. 4, pp.
607–618, Dec. 2007.
Book:
[5] (Bradley Lee Barnhart University of
Iowa,2011) the
hilbert huang transform
applications development .
693 | P a g e
4. Sonali Dhanorkar et al Int. Journal of Engineering Research and Applications
ISSN : 2248-9622, Vol. 4, Issue 2( Version 1), February 2014, pp.691-696
www.ijera.com
Ist Image sequences
(a)
(b)
(c)
(d)
(e)
Fig.4.(a) Vertical LMV and Horizontal LMV Signal, (b)Horizontal (X) LMV Signal ,upper and lower envelopes and mean signal
,(c)generated IMF and .(d) vertical (Y) LMV Signal ,upper and lower envelopes and mean signal (e) Produced IMF.
IInd Image sequences
www.ijera.com
694 | P a g e
5. Sonali Dhanorkar et al Int. Journal of Engineering Research and Applications
ISSN : 2248-9622, Vol. 4, Issue 2( Version 1), February 2014, pp.691-696
www.ijera.com
(a)
(b)
(c)
(d)
(e)
Fig.5.(a) Vertical LMV and Horizontal LMV Signal,(b)Horizontal(X) LMV Signal ,upper and lower envelopes and mean signal,
(c) generated IMF ,(d) Vertical LMV(Y) Signal ,upper and lower envelopes and mean signal and (e) generated IMF
www.ijera.com
695 | P a g e
6. Sonali Dhanorkar et al Int. Journal of Engineering Research and Applications
ISSN : 2248-9622, Vol. 4, Issue 2( Version 1), February 2014, pp.691-696
www.ijera.com
IIIrd Image sequences
(a)
(b)
(c)
(d)
(e)
Fig.6.(a) Vertical LMV and Horizontal LMV Signal,(b)Horizontal(X) LMV Signal ,upper and lower envelopes and mean signal,
(c) generated IMF ,(d) Vertical LMV(Y) Signal ,upper and lower envelopes and mean signal and (e) generated IMF
www.ijera.com
696 | P a g e