Reversible Data Embedding Using a Difference Expansion Speaker: Meng-Jing Tsai Author: Jun Tian Source: IEEE Transactions on Circuits and Systems for Video Technology, Vol. 13, No. 8, Aug. Date: 2012.05.3
Feb 22, 2016
Reversible Data Embedding Using a Difference Expansion
Speaker: Meng-Jing TsaiAuthor: Jun TianSource: IEEE Transactions on Circuits and Systems for Video Technology, Vol. 13, No. 8, Aug. 2003, page(s ): 890-896
Date: 2012.05.31
2
Outline
• Introductions– Reversible Hiding– Difference Expansion
• The Proposed Method• Experimental Results• Conclusions
3
Reversible Hiding
• It can remove the embedded data to restore the original image.
embedding extracting
secret data
restore the original image
secret data
4
Difference Expansion
201 206mean = 203
h = 206-201 =5 = (101)2
(101b)2 = (1011)2 = 1165
198 209mean = 203
-5 +6
expansionb=1
h’ =209-198=2×h+b=11
21hmean
2hmean
pixel pair (201,206) → (198,209)
5
Data Embedding Difference Expansion Method
(x,y)h=x-y
Case1. expandable : 2×h+bCase2. changeable : LSB(h)Case3. non-changeable
bitstream C
location map L
100
6
Case1. Expandable
201 206mean = 203
h = 206-201 =5 = (101)2
(101b)2 = (1011)2 = 1165
198 209mean = 203
-5 +6
expansionb=1
≦255 ≦255
location map L=1
h’ =209-198=2×h+b=11
7
≦255 ≧ 255
Case2. Changeable200 246
mean = 223
h = 246-200 =46 = (101110)2
(101110b)2 = (1011101)2 = 934746
177 270mean = 223
-46 +47
expansionb=1
b=1 (101110)2 = (101111)2 = 472423
200 247mean = 223
-23 +24≦255 ≦255
location map L=0
=2× +bh’ =247-200
=47
h-└ ┘2
8
Expandable Decodinglocation map L=1
198 209mean = 203
h’ =209-198=11=(1011)2
=(101b)2
b=1h=5
x=206y=201
23
9
Changeable Decoding
200 246mean = 223
h’=247-200= 47=(101111)2
=(10111b)2
h=(101110)2
= 46
b=1h=( 101110)2 =46
x=246y=200
location map L=0
65
10
Multiple-Layer Embedding
201 206mean = 203
h = 206-201 =5 = (101)2
(101b)2 = (1011)2 = 1165
198 209mean = 203
-5 +6
expansionb=1
Layer1
Layer2 h = 209-198 =11 = (1011)2
(1011b)2 = (10110)2 = 221111
192 214mean = 203
-11 +11
expansionb=0
“Lena,” with a 39566 bits (0.15 bpp) payload. “Lena,” with a 516794 bits (1.97 bpp) payload.
Experimental Results
“Lena,” with a 141493 bits (0.54 bpp) payload.11
12
Conclusions• It’s a simple and efficient reversible data
embedding method for digital images.
• It achieves very high payload capacity.
• It keeps the distortion low.
13
Thank you for your listening.