1 Adaptive Data Hiding in Edge Areas of Images With Spatial LSB Domain Systems Adviser: Chih-Hung Lin Speaker ： Chia-Wei Chang Date:2009/10/06.

1

Adaptive Data Hiding in Edge Areas of Images

With Spatial LSB Domain Systems

Adviser Chih-Hung LinSpeaker Chia-Wei ChangDate20091006

2

Introduce of Authorsbull AuthorCheng-Hsing Yang Chi-Yao Weng Shiuh-Jeng Wang and Hung-

Min Sun

bull IEEE Transactions on Information Forensics and Security vol 3 no 3 september 2008

bull C-H Yang is with the Department of Computer Science National Pingtung University of Education Pingtung 900 Taiwan

bull C-Y Weng and H M Sun are with the Department of Computer Science National Tsing-Hua University Hsinchu 300 Taiwan ROC

bull S-J Wang corresponding author is with the Department of Information Management Central Police University Taoyuan 333 Taiwan ROC

bull Manuscript received November 1 2007 revised March 12 2008 Published August 13 2008 (projected)

3

Outline

bull Introductionbull Literature Reviewbull Proposed Method

Embedding Procedure Data Extraction

bull Experimental Resultsbull Conclusions

4

IIntroduction

5

IIntroduction

embedding

=

bullEmbedding

6

IIntroductionbull In 2003 Wu and Tsai proposed a ldquoPVDrdquo

bull In2004 Chang and Tseng proposed a side-match method

bull In 2005 Park et al proposed a steganographic scheme based of neighboring pixels

bull In 2005 Wu et al proposed the pixel-value differencing (PVD) and LSB

replacement method

7

IIntroduction

bull LSB techniquesbull PVD techniquesbull Side-Match techniques

8

LSB method

Fig1Wu et alrsquos division of ldquolower levelrdquo and ldquohigher levelrdquo (Div=15)1048576

9

II Literature Reviewbull Wu et alrsquos PVD and LSB

START

nonoverlappingblocks with two

consecutive pixels

belongs to the higher levelid belongs to

the lower levelid

Emdedded by the 3-b LSB

Emdedded by the 3-b LSB

)()( 1

1

iiii pppp)()( 1

1

iiii pppp

Divdifppd

i

iii

|| 1

= =

1

1

1

1

)88()88(

1

)(

iiii

iiii

ppifppppifppii pp

YN

NN

END

YY

leveld

leveldif

i

i

leveld

leveldif

i

i

10

II Literature Reviewbull Example

)2(1 11000sec475615 bdataretandppDiv ii

id =|56-47|=9 (lower-level)

To embed 111000ip

1ip=63 =40

id =23gtDiv=15

1

1

1

1

)88()88(

1

)(

iiii

iiii

ppifppppifppii pp

If levelne levelidid

ip

1ip=55 =48

id =55-48=7

11

IIIProposed Method

bull 2-3 Division and 3-4-5 Division

(a)

(b)

Fig 2 Two dividing cases (a) ldquolower levelrdquo and ldquohigher levelrdquo and (b) ldquolowerlevelrdquo ldquomiddle levelrdquo and ldquohigher levelrdquo

12

IIIProposed Method

bull 2-3 Division

bull 3-4-5 Division

||log||log 2212 RhandRlhl

||log||log||log 322212 RhandRmRlhml

13

III-1Embedding steps for l-m-h

14

III-1Embedding steps for l-m-h

15

III-1Embedding steps for l-m-hbull Case61 =L-level neL-level

bull Case62 =M-level =L-level

bull Case63 =M-level =H-level

bull Csae64 =H-level neH-level

)2()2()(

)2()2()(

1

1

1

1

1

1

1

1

ik

ik

iiiiii

ik

ik

iiiiii

pporpppppp

pporpppppp

)2()2()(

)2()2()(

1

1

1

1

1

1

1

1

ik

ik

iiiiii

kiii

kiiiii

pporpppppp

pporpppppp

)2()2()(

)2()2()(

1

1

1

1

1

1

1

1

ik

ik

iiiiii

ik

ik

iiiiii

pporpppppp

pporpppppp

)2()2()(

)2()2()(

1

1

1

1

1

1

1

1

ik

ik

iiiiii

ik

ik

iiiiii

pporpppppp

pporpppppp

id

id

id

id

id

id

id

id

16

III-1 Embedding Procedure Example

)2(1 10100000sec476415 bdataretandppDiv ii

id =|64-47|=17 (middle-level)

To embed 4-bit LSBip

1ip=74 =32

id =42

=M-level =L-levelidid

ip

1ip=74 =48

id =|74-48|=26

After the modified LSB substitution is appliedip

1ip=58 =48

id =10

)2()2()(

)2()2()(

1

1

1

1

1

1

1

1

ik

ik

iiiiii

kiii

kiiiii

pporpppppp

pporpppppp

17

III-2Data Extraction

18

There are 4 bits embedded in and 4 bits embedded in

III-2Data Extraction Exampleip

1ip=74 =48

id =|74-48|=26

ip

1ip

lsquos secret bits

lsquos secret bits )2(0000

ip

)2(1010

1ip

19

IVExperimental Results

Fig 3Two cover images

(a)Elaine (b) Baboon

20

IVExperimental Results

21

IVExperimental Results

22

IVExperimental Results

23

IVExperimental Results

24

VConclusions

bull some l-h divisions and l-m-h divisions both yielded higher capacity and higher PSNR

bull Embed secret data into gray images without making a perceptible distortion

bull Can be analyzed automatically aiming to capture an adequate - - division satisfying the key requirements in information hiding

bull Experimental results show that our approach obtains both larger capacity and higher image quality than that of Wu et al