Geometric wavelet transform for optical flow estimation algorithm

International Journal of Computer Graphics & Animation (IJCGA) Vol.4, No.2, April 2014
DOI : 10.5121/ijcga.2014.4201 1
GEOMETRIC WAVELET TRANSFORM FOR OPTICAL
FLOW ESTIMATION ALGORITHM
Cheriet Leila1
, Chenikher Salah2
and Boukari karima3
1
Department of Electronic, Annaba University, Annaba, Algeria
2
Department of Genie Electric, Tebessa University, Tebessa, Algeria
3
Department of Electronic, Annaba University, Annaba, Algeria
ABSTRACT
This paper described an algorithm for computing the optical flow (OF) vector of a moving objet in a video
sequence based on geometric wavelet transform (GWT). This method tries to calculate the motion between
two successive frames by using a GWT. It consists to project the OF vectors on a basis of geometric
wavelet. Using GWT for OF estimation has been attracting much attention. This approach takes advantage
of the geometric wavelet filter property and requires only two frames. This algorithm is fast and able to
estimate the OF with a low-complexity. The technique is suitable for video compression, and can be used
for stereo vision and image registration.
KEYWORDS
Geometric Wavelet, Curvelet Wavelet, Motion Estimation, Optical Flow
1. INTRODUCTION
Recherch work on optical flow estimation has previously been approved out to different
applications which found in various fields such as signal and image processing; pattern
recognition and computer vision, astronomy, acoustics and geophysics. It also finds place in
medicine and meteorology by processing related images. Another area of high importance is the
detection and tracking of moving targets in military applications [8],[9],[10] and [12].
A Great number of approaches for OF estimation have been proposed in the literature, including
gradient-based, correlation-based, energy based, and phase based techniques [1].
The use of real wavelet in this approach suffers from two main problems; the lacks of shift
invariance and poor directional selectivity. The geometric wavelet solves these problems and still
desired to provide special characteristics.
The wavelet curvature is applicable principally in motion estimation edge detection and texture
discrimination.

2
The rest of the paper is organized as follows. In Section 2, we introduce the principle and the
description of the proposed algorithm. Section 3 shows the experimental performance of optical
flow estimation. Finally, Section 4 concludes our contribution and merits of this work.
2. OPTICAL FLOW ESTIMATION ALGORITHM
2.1. Geometric Wavelet
Curvelet transform correspond on association of different steps: the application of filters pass-
bands, on segmentation dyadic of each band frequency and transform Ridgelet on each zone
segmented. The decomposition frequented associated on the dyadic segmentation permit
conditioner the data for the Ridgelet transform with the aim of describe the singularities, where
the size and the form of the motif variant. The description of the singularities where the size and
the form of the motif variant. The description correspond the variability on the position: position
associates on different dyadic zones.
The aim of this variant is permit the separation in the signal where the motif is very complex that
in Ridgelet transform. This transform is very adapted for the split of a real image, it approximate
the contour from an ensemble of segments. This transform ameliorant the singularities of an
image, on the contrary, it is very redundant. It exists the alternatives for the limitation of the
redundancy.
Figure (1): (a) Decomposition of an image with filter banc. (b) Application of Ridgelet transform
in each zone dyadic.
2.2. Gradient Constraint
Our work shows that geometric wavelet transform technique seems more accurate in optical flow
estimation.
This technique is based on the assumption of the gradient constraint. This equation can be
expressed as
( ) ( )

3
Or
Which result in:
Thus:
⃗
Where , are the optical flow of ( ) and , and or , and are the derivatives
spatial and temporal of the image.
Our method introduces an additional condition for estimating the optical flow because it is an
equation in two unknowns and cannot be solved as such. To discover the optical flow another set
of equations is needed, given by some additional constraint.
2.3 Method
The framework of the algorithm is illustrated:
Where
By applying the geometric wavelet transforms at the level l, we get the hierarchical gradient
constraint functions (see figure 1).
The global equation of gradient constraint at all scale levels L is:
tX IpYXfI ),(.
,
1000
0001
),(
T
iyix
iyix
YXf


























MM yx
yx
X
II
II
I



0000
00
000011
l
t
l
IpA 
bAp 
],,,[ 10 L
AAAA  TL
ttt IIIb ),,( 10


4
3 Experiment results:
In our experiments, we used three of sequences synthetic and four methods for comparison. For
our simulation the low and high pass filter are done by
We evaluate the optical flow by using the angular error measurement between the correct velocity
( ) and the estimate velocity ( ) with density, the average error and standard
deviation were calculated.
Three images sequences were used to test our algorithm and compared with other optical flow
technique:
(
√ √
)
Figure 2.A. Sequence mysineB-6
Figure 2.B. Optical Flow measuring by using real wavelet transforms

5
Figure 2C. Optical Flow measuring by using geometric wavelet transforms
Figure 3.A. Sequence Diverging tree
Figure 3C. Optical Flow measuring by using Geometric wavelet transforms

6
Figure 4.A. Sequence Yosemite
Figure 4C. Optical Flow measuring by using Geometric wavelet transforms
Table 1.Comparison of different methods of sequence “my sineB.6”.
Method Images Error deviation Density (%)
Motion estimation using complex
wavelet
2 0.79° 0.68 100
Motion estimation using real wavelet 2 5.35° 4.56° 100
Horn et Schunk (original) 2 12.02° 11.72° 100
Horn et Schunk (modify) 7-13 2.55° 3.67° 100
Anandan 2 7.64° 4.96° 100
Singh 2 8.60° 4.78° 100

7
Table 2.Comparison of different methods of sequence “Diverging tree”.
Motion estimation using complex wavelet 2 0.79° 0.68 100
Horn et Schunk (original) 2 12.02° 11.72° 100
Anandan 2 7.64° 4.96° 100
Singh 2 8.60° 4.78° 100
Table 3.Comparison of different methods of sequence “Yosemite”.
Motion estimation using complex
wavelet
2 4.52° 4.65° 100
Horn et Schunk (original) 2 32.43 ° 30.28° 100
Anandan 2 15.84° 13.46° 100
Singh 2 13.16° 12.07° 100
4. Conclusion
The algorithm estimate the OF by resolving the resulted equations from the projection of the
gradient constraint in the Geometric wavelet bases. The experimental results demonstrate that the
Geometric wavelet is capable to estimate many kinds of movement. Testing on synthetic data sets
has shown good performance compared par real wavelet.
REFERENCES
[1] Lee, S.hyun. & Kim Mi Na, (2008) “This is my paper”, ABC Transactions on ECE, Vol. 10, No. 5,
pp120-122.
[2] Gizem, Aksahya & Ayese, Ozcan (2009) Coomunications & Networks, Network Books, ABC
Publishers.
genetic trail bounded approximation for H. 264/AVC codecs- MA Haque - Multimedia Tools and
Applications, 2012 – Springer
[1] Models for Local Optical Flow Estimation- T Corpetti – Scale Space and Variational Methods in
Computer, 2012 – Springer.
[2] S.LEE, Fast motion estimation based on search range adjustment and matching point decimation,
Department of Imaging Engineering, ChungAng University Seoul, COREE, REPUBLIQUE DE,
2010.
[3] Large Displacement Optical Flow: Descriptor Matching in Variational Motion Estimation Brox,
[4] T.; Malik, J.; Dept. of Electr. Eng Comput. Sci.,Univ. of California at Berkeley, Berkeley, CA, USA-
2011

8
[5] A.Ahmadi, M.M.Azadfar, “Implementation of fast motion estimation algorithms comparison with
full search method in H.264”, IJCSNS International Journal of Computer Science and Network
Security, VOL.8 No.3, March 2008
[6] J.Barron, D.Fleet, and S. Beauchemin.“Performance of optical flow techniques”. Int. J. Computer
Vision, 1994.
[7] W Lawton, „Applications of Complex Valued Wavelet Transforms to Subband Decomposition‟, IEEE
Trans. Sig. Proc., 41, 12, 3566-3568, 1993.
[8] N G Kingsbury, „Image Processing with Complex Wavelets‟, Phil. Trans. Royal Society London,
1999

Geometric wavelet transform for optical flow estimation algorithm

More Related Content

What's hot (20)

Similar to Geometric wavelet transform for optical flow estimation algorithm (20)

Recently uploaded (20)

Geometric wavelet transform for optical flow estimation algorithm