Вы находитесь на странице: 1из 4

2008 International Conference on Computer Science and Software Engineering

A Novel Steganographic Algorithm Based on the


Motion Vector Phase

Xuansen He Zhun Luo


College of computer and communication College of computer and communication
Hunan University Hunan University
Changsha, China Changsha, China
hxshf@vip.sina.com gol_ele@sina.com

Abstract—Most data hiding techniques in digital video utilize I Changyong Xu et al propose that the secret data are embedded
frame to embed the secret information so the capacity of P and B in the amplitude of the motion vector and the control data are
frame is wasted. In this paper, we first analyze a data-hiding embedded in I frame [2]. Dingyu Fang et al proposes a
algorithm using the phase angle difference of the motion vector, data-hiding scheme that embeds the secret data in digital videos
on this base a novel steganographic algorithm based on the phase using the phase angle difference of the motion vector[3]. Shan
of the motion vector is proposed. The algorithm utilizes the phase Zhu et al proposes the diamond search algorithm for fast
of single motion vector to embed the secret data in P or B frame, block-matching motion estimation[4]. Above the methods have
in order to improve the embedding efficiency we use the matrix the advantages of least influence on the videos and simple
encoding technique to achieve a better tradeoff between the
algorithms. In this paper we proposes an new effective
quantities of hidden data and the motion vectors modification
rate. The simulation results have validated the feasibility of the
algorithm that embeds the secret information into P or B frame
algorithm. of the phase of the motion vector. The remainder of this paper
is organized as follows. Section 2 introduce a data hiding
Keywords-seganography; phase of motion vector; matrix algorithm based on the phase angle difference of the motion
encoding; digital video vector. Section 3 is dedicated to deriving a new seganographic
algorithm based on the phase of the motion vector. The
I. INTRODUCTION simulation results of the algorithms are presented in Section 4.
Data hiding technique (seganography or information Finally, Section 5 concludes the paper.
hiding) has become an important research content in the field II. THE ALGORITHM BASED ON THE PHASE ANGLE
of information security. The up-to-date information hiding is
DIFFERENCE OF THE MOTION VECTOR
that the important data or secret information is embedded into
some media such as images, videos or audios without The foundation of the data hiding algorithm based on the
introducing perceptual distortion. After receiving the media phase angle difference of the motion vector [3] (method _first)
embedded the secret information, the receiver extracts the is as follows: most of the video encoders perform some fast
information by the corresponding key and algorithm. Because block search algorithm to substitute the full search algorithm,
the information quantities of uncompressed video are big so the these fast search algorithms include three steps search (TSS),
videos have to be compressed in order to save transmission four steps search (FSS) and diamond search(DS) [4],[5], and so
bandwidth and storage space. The most famous compressing on, these algorithms can find the local optimal motion vector
and coding standards are MPEG-X (X=1,2,4) and H.26X and can save lots of computational burden. The method_first
(X=1,2,3,4), the information hiding algorithms on those replaces the original motion vector with another local optimal
standards are called digital video or compressed video motion vector to embed data in the motion vectors in the inter
algorithms. The digital video has become an important frame, the process of the algorithm is as below [3]:
information hiding medium because of its big signal capacity (1) Discard the surrounding macroblock in the current inter
and extended applications range. Up to now, there are many frame.
techniques for embedding the secret information into digital
video. The steganography in digital video can choose different (2) Compute the motion vector magnitudes in the
position to embed the secret information, for example, the remaining macroblocks.
secret information is added to the DCT coefficients, the
(3) Using the predefine threshold T to select a set S for
information is embedded in the motion vectors, etc.
candidate motion vectors(MV).
The algorithms based on the motion vectors select the
appropriate motion vector first and then embed the secret S = {MV0 , MV1 , " , MVn −1} , |S|=n
information into the amplitude or phase. Jordan et al propose
information hiding in digital video using INTER-frame [1]. ( MViV2 + MViH2 ) ≥ T (0 ≤ i < n) , where MViH and MViV is
They select the inter frames in MPEG video sequences, and respectively the horizontal and vertical components of the
regularize the motion vectors into a modified sequences.

978-0-7695-3336-0/08 $25.00 © 2008 IEEE 822


DOI 10.1109/CSSE.2008.359
motion vector MVi . The above two sectors data hiding algorithm for embedding
(4) Compute the phase angle of each motion vector in S: 1 bit in the phase angle of a pair of motion vectors can be
extended to multi-sectors. The circle can be segmented into
Φ i = arctan ( MViV MViH ) more sectors and every sector denotes different embedded
bits respectively. For example, four sectors (0°,90°),
(5) Get a pair of motion vectors MV2i and MV2i +1
(90°,180°), (180°,270°), (270°,360°), where (0°,90°) and
(0≤i≤[n/2]) from S in a sequential order. (180°,270°) attribute to one sector set. Reference [3] has done
(a) If data bit is 0, the two conditions are considered: the experiment and proved that the sixteen sectors improves
PSNR approximate 2.5dB comparing with the two sectors.
i) If 0° <| Φ 2i − Φ 2i +1 |≤ 180° the both motion vectors are
III. THE ALGORITHM BASED ON THE PHASE OF THE
not changed.
MOTION VECTOR
ii) If 180° <| Φ 2i − Φ 2i +1 |≤ 360° the new motion vectors There are two disadvantages in the method_first: one is that
MV2∗i and MV2∗i +1 are searched for the (2i)-th and (2i+1)-th 1 bit secret data is embedded in two motion vectors so the
utilizing efficiency of the media is lower, another is that the
macroblocks such that | MV2∗i |≥ T , | MV2∗i +1 |≥ T ,
computation and replacing amount of the motion vectors is big.
∗ ∗
0° <| Φ − Φ 2i +1 |≤ 180° and 0° <| Φ 2i − Φ
2i 2 i +1 |≤ 180° . In order to solve the above disadvantages we proposes an
algorithm based on the phase of the motion vector
Let MSE2i = MSE between B∗p 2i and B p 2i , where the (method_second).
prediction block Bp2i with MV2i in the reference frame, We also segment circle (0°, 360°) into two sectors and the
prediction block B∗p 2i with MV2i∗ in the reference frame. Let two sectors express respectively data “0” and “1”, either
MSE2i +1 = MSE between B∗p (2i +1) and B p (2i +1) , where the motion vector must be attributed to one of the two sectors.
When the embedded data match with the phase information of
prediction block B p (2i +1) with MV2i +1 in the reference frame, the motion vector, then the motion vector is not changed,
the prediction block B∗p (2i +1) with MV2∗i +1 in the reference otherwise, we search one optimal motion vector in another
sector to replace the original motion vector. The method_first
frame. If (MSE2i<MSE2i+1) then MV2i is replaced with MV2i∗ , and method_second respectively adopt different set of
otherwise, MV2i+1 is replaced with MV2∗i +1 . coordinates. The vertical coordinate axis of the method_first is
the motion vector that is not changed, but the method_second
(b) If data bit is 1, the two conditions are considered: uses fixed right-angled coordinate system. That is showed in
Figure 1.
i) If 180° <| Φ 2i − Φ 2i +1 |≤ 360° , the both motion vectors
are not changed. When MV2 needs to be changed, the two algorithms only
ii) If 0° <| Φ 2i − Φ 2i +1 |≤ 180° , the new motion vectors search an optimal MV2∗ in the shadow section of Figure 1 to
substitute MV2 . Our algorithm can improve the utilizing
MV2i∗ and MV2∗i +1 are searched for the (2i)-th and (2i+1)-th
efficiency of the media but cannot solve the problem that the
macroblocks such that | MV2∗i |≥ T and | MV2∗i +1 |≥ T , motion vectors modification amount is too big, so we introduce
∗ ∗
180° <| Φ − Φ 2i +1 |≤ 360° ,
2i 180° <| Φ 2i − Φ 2 i +1 |≤ 360° . matrix encoding to improve the performance of the algorithm.
Matrix encoding is a circulation encoding technique, it can
Let MSE2i = MSE between B∗p 2i and B p 2i , where the
improve the information embedding efficiency. Per three
prediction block B p 2i with MV2i in the reference frame, sequential motion vectors that are satisfied with the amplitude
prediction block B∗p 2i with MV2i∗ in the reference frame. Let conditions are segmented into a group, and 2 bits secret
information are embedded in per group. We only modify one
MSE2i +1 = MSE between B∗p (2i +1) and B p (2i +1) , where the motion vector once, for example, MV1 , MV2 , MV3 are in
prediction block B p (2i +1) with MV2i +1 in the reference frame, one group, if MV1 and MV3 need to be modified
the prediction block B∗p (2i +1) with MV2∗i +1 in the reference synchronously, we only modify MV2 to satisfy the
frame. If (MSE2i<MSE2i+1) then MV2i is replaced with MV2i∗ , corresponding relation.
otherwise, MV2i+1 is replaced with MV2∗i +1 . In the method_first, if MV2 needs to be modified but
MV1 and MV3 are fixedness, then the phase of MV2 must
(6) If the embedding process is not completed, then go to
Step (5), otherwise, go to Step (1) to process the next inter be satisfied with the two conditions of MV1 and MV3
frame. synchronously, this method restricts the searching area and
makes the difference between the optimal vector and the
(7) The extraction process of the secret information: if original vector increase. The matrix encoding of the
0° <| Φ 2i − Φ 2i +1 |≤ 180° then the secret data bit is 0, if method_first in four sectors is showed in Figure 2. From the
180° <| Φ 2i − Φ 2i +1 |≤ 360° then the data bit is 1. figure we can obviously see that the searching area has been
decreased.

823
i) If Q1⊕Q2=0 and Q2⊕Q3=0, then MVi + 2 is modified
into another phase sector.
ii) If Q1⊕Q2=0 and Q2⊕Q3=1, then every motion vectors
are not changed.
iii) If Q1⊕Q2=1 and Q2⊕Q3=0, then MVi +1 is modified
into another phase sector.
iv) If Q1⊕Q2=1 and Q2⊕Q3=1, then MVi is modified
(a) The method_first into another phase sector.
(c) If embedded data are “10”, we consider four conditions
as below:
i) If Q1⊕Q2=0 and Q2⊕Q3=0, then MVi is modified into
another phase sector.
ii) If Q1⊕Q2=1 and Q2⊕Q3=0, then every motion vectors
are not changed.
iii) If Q1⊕Q2=0 and Q2⊕Q3=1, then MVi +1 is modified
(b) The method_second
into another phase sector.
Figure 1. The set of coordinates of the two algorithms
iv) If Q1⊕Q2=1 and Q2⊕Q3=1, then MVi + 2 is modified
into another phase sector.
(d) If embedded data are “11”, we consider four conditions
as below:
i) If Q1⊕Q2=0 and Q2⊕Q3=0, then MVi +1 is modified
into another phase sector.
ii) If Q1⊕Q2=1 and Q2⊕Q3=0, then MVi + 2 is modified
into another phase sector.

Figure 2. The matrix encoding of the method_first iii) If Q1⊕Q2=0 and Q2⊕Q3=1, then MVi is modified into
The process of our algorithm can be written as follows. another phase sector.
Step (1) to Step (4) are the same with that of the iv) If Q1⊕Q2=1 and Q2⊕Q3=1, then every motion vectors
method_first. are not changed.
(5) The sequential three motion vectors MVi, MVi+1, MVi+2 (6) If the embedding process is not completed, then go to
in S are segmented into a group. Q1, Q2, Q3 express Step (5), otherwise, go to Step (1) to process the next inter
respectively the embedded data of motion vectors MVi, MVi+1, frame.
MVi+2, so the value of Qi is “0” or “1”.
(7) The extraction of the secret information: we only
i) If Q1⊕Q2=0 and Q2⊕Q3=0, where sign “⊕” denotes the compute the value of Q1⊕Q2, Q2⊕Q3 and can easily get the
“exclusive or” operator, then every motion vectors are not embedded secret information.
changed.
VI. EXPERIMENT RESULTS ANALYSIS
ii) If Q1⊕Q2=1 and Q2⊕Q3=0, then MVi is modified into To test the performance of the algorithms, we make a series
another phase sector. of experiments. In the experiment, the two methods are applied
to the MPEG-2 digital video stream and we segment circle (0°,
iii) If Q1⊕Q2=0 and Q2⊕Q3=1, then MVi + 2 is modified 360°) into sixteen sectors. For our algorithm we embed a
into another phase sector. random sequence of the secret information into P frame of the
tested video and choose “IPPP” encoding.
iv) If Q1⊕Q2=1 and Q2⊕Q3=1, then MVi +1 is modified
In the simulations we compare the performances of the two
into another phase sector. algorithms, and the same simulation conditions are used for the
(b) If embedded data are “01”, we consider four conditions two algorithms to allow fair comparison.
as below:

824
In the first simulation we compare the embedding capacity simulation result is showed in Figure 5. From it we can see that
of the two algorithms. The simulation conditions are T=10 and our algorithm can improve PSNR about 2dB.
the video signals are the sequential eight P frames of Stefan
video. The result of this simulation is showed in Figure 3.
From it we can see that our algorithm improves the embedding
capacity about 15%.

Figure 5. The PSNR comparison

V. CONCLUSION
We propose a new steganographic algorithm based on the
phase of the motion vector in this paper. In the algorithm the
Figure 3. The embedding capacity comparison sequential three motion vectors are segmented into a group
In the second simulation we compare the motion vectors and 2 bits data are embedded in per group, once only one
modification rate of the two algorithms, we define the motion
motion vector is modified at most so that the MV modification
vectors (MV) modification rate as below:
rate is decreased and the embedding efficiency is improved.
the number of modified MV Our algorithm can be applied to compressed videos such as
MV modification rate =
the number of embedded data bits MPEG-X (X=1, 2, 4) or H.26X (X=1, 2, 3, 4) and can be
compatible with other types of data hiding algorithms. The
The test conditions are 20 groups random sequences of experimental results show that our algorithm not only can
Stefan video. The simulation result is showed in Figure 4. embed large amounts of the secret information into a video but
From it we can see that our algorithm can decrease the motion
also can maintain good video quality.
vectors modification rate about 10%.
ACKNOWLEDGMENT
This work was supported by the Key Item of Science and
Technology Program of Xiangtan City, Hunan Province, China
(No. ZJ20071008).
REFERENCES
[1] Kutter M, Jordan F, Ebrahimi T. Proposal of a watermarking technique
for hiding/retrieving data in compressed and decompressed video [R].
StockhoLm: ISO/IEC, 1997
[2] Xu Changyong, Ping Xijian, Zhang Tao. Steganography in Compressed
Video Stream. Proc of International Conference on Innovative
Computing, Information and Control[C]. China. IEEE Society Press,
2006: 269-272
Figure 4. The motion vector modification rate comparison [3] Fang Dingyu, Chang Longwen. Data Hiding For Digital Video with
Phase of Motion Vector. Proc of International Symposium on Circuit
Because the two algorithms are materially the same in and Systems (ISCAS)[C]. Greece. IEEE Society Press, 2006:1422-1425
modifying the single motion vector, then the lower MV [4] Zhu S, Mak K. A new diamond search algorithm for fast block matching
modification rate can directly improve the quality of the video motion estimation[J]. IEEE Trans image Process,2000,9(2):287—290
embedded the secret information. [5] Cheung C H, Po L M. Novel cross-diamond-hexagonal search
algorithms for fast block motion estimation[J]. IEEE Trans
In the third experiment we compare the signal-to-noise ratio Multimedia,2005,7(1):l6—22
(PSNR) of the two algorithms. The test conditions are Football
video sequences. Here we chiefly compare the relation
between the PSNR and the number of embedded data bits. The

825

Вам также может понравиться