The Study of Vector Quantization and Its Applications to Information Hiding ???????????????????

1
The Study of Vector Quantization and Its
Applications to Information Hiding??????????????
?????

• Advisor Chin-Chen Chang1, 2
• Student Wen-Chuan Wu2
• 1 Dept. of Information Engineering and
  Computer Science,
• Feng Chia University
• 2 Dept. of Computer Science and Information
  Engineering,
• National Chung Cheng University

2
The Fields of Vector Quantization

• VQ" Block-based quantizer
• Applications
• Signal compression (i.e. Image, Speech, )
• Feature recognition
• Information security
• Video-based event detection
• Anomaly intrusion detection

3
Outline

• Part I Design and Analysis of VQ-
  Based Algorithms
• Part II VQ Applications to Information
  Hiding
• Fixed Embedding
• Adaptive Embedding
• Reversible Embedding

4
Part I Design and Analysis of VQ-Based
Algorithms

• Fast Planar-Oriented Ripple Search Algorithm for
  Hyperspace VQ Codebook

5
VQ

• Overview

X
512
6

• Related works
• Full-search equivalents
• Rough distortion elimination to filter impossible
  codewords
• Partial-search methods
• Organize the codebook by some data structures to
  label a local search domain (Array, Binary Tree,
  )

7

• Comparison
• Full-search equivalents
• To get the best result.
• To consume much computation time on-line for a
  rejection test.
• Partial-search methods
• Some operations are off-line. (Fast searching)
• To need extra memory space.
• To get a closer result.

8
Planar-Oriented Ripple Search(Planar Voronoi
Diagram Search PVDS 15)

• How to construct a Voronoi diagram

Perpendicular bisector
9
Planar-Oriented Ripple Search(Planar Voronoi
Diagram Search PVDS 15)
An example of planar Voronoi diagram with 13
points
10

• Experiments (Without LUT operation)

RipplesImages RipplesImages 0 1 2 3 4 5 DTPC FSVQ
Lena PSNR (dB) 26.910 30.920 31.254 31.313 31.334 31.336 30.623 31.338
Lena Time (sec.) 0.111 0.233 0.623 1.298 2.140 3.687 0.375 9.937
Lena Count 1 6.9 19.2 37.5 62.4 95.2 5.7 256
Boat PSNR (dB) 24.990 28.853 29.227 29.295 29.308 29.314 28.438 29.321
Boat Time (sec.) 0.142 0.265 0.597 1.232 2.173 3.766 0.391 9.922
Boat Count 1 6.8 18.8 37.3 62.9 95.1 6.1 256
Jet PSNR (dB) 25.797 30.085 30.438 30.504 30.513 30.515 29.529 30.517
Jet Time (sec.) 0.141 0.250 0.531 1.141 1.936 3.332 0.343 9.891
Jet Count 1 6.3 17.9 34.7 56.3 86.4 5.1 256
Codebook size 256
11

• Experiments (Without duplication)

MethodsImages MethodsImages TSVQ25 CPTSVQ 4 SCS51 DTPC8 HOSM56 PVDS15 FSVQ
Lena PSNR (dB) 28.238 29.225 30.658 30.623 30.005 31.254 31.338
Lena Time (sec.) 1.656 1.797 0.647 0.375 0.687 0.623 9.937
Lena Count 52 56 17.2 5.7 17.3 19.2 256
Boat PSNR (dB) 26.539 27.436 28.493 28.438 28.246 29.227 29.321
Boat Time (sec.) 1.656 1.781 0.619 0.391 0.671 0.597 9.922
Boat Count 52 56 11.8 6.1 17.2 18.8 256
Jet PSNR (dB) 27.532 28.494 29.671 29.529 29.448 30.438 30.517
Jet Time (sec.) 1.656 1.797 0.621 0.343 0.640 0.531 9.891
Jet Count 52 56 12.0 5.1 16.1 17.9 256
Codebook size 256
12

• Experiments (With duplication)

MethodsImages MethodsImages SCS51 HOSM56 PVDS15 FSVQ
Lena PSNR (dB) 31.338 31.338 31.338 31.338
Lena Time (sec.) 0.971 1.448 0.910 9.937
Lena Count 28.2 38.5 29.2 256
Pepper PSNR (dB) 30.596 30.707 30.708 30.712
Pepper Time (sec.) 0.866 1.433 0.950 9.890
Pepper Count 25.0 37.8 27.5 256
Baboon PSNR (dB) 24.038 24.068 24.078 24.104
Baboon Time (sec.) 0.971 1.501 1.278 9.907
Baboon Count 28.3 39.4 32.9 256
Codebook size 256
13

• Experiments

Methods TSVQ CPTSVQ SCS DTPC HOSM PVDS FSVQ
Encoding complexity O(k log2N) O(k log2N) O(log2GD NG) O(THk log2N) O(log4N NH) O(log2N NV) O(kN)
Duplication ability No No Yes No Yes Yes No
14
Part II VQ Applications to Information Hiding

• Hiding secrets in VQ (SMVQ) codes
• Adaptive embedding
• Reversible data hiding

15
Data Hiding
Compressed codes 10111011 11..
11010110 01..
Information
(187)10
(214)10
Internet
Sender
Receiver
Information
16
Data Hiding in VQ Codes
(Jo and Kim 29)
18
18
46
46
19
17
Side Match VQ (SMVQ)

• Assumption Neighboring pixel intensities in an
  image are pretty similar.

18
X (81, 15, 53, 34, 51,?, ?, ?, 91, ?, ?, ?,
49,?, ?, ?)
Codebook(512)
State codebook(16)
19
Data Hiding in SMVQ and VQ Codes
12
1
1
THSMVQ
THVQ
Bit1
20

• Experiments

MethodsImages Jo and Kims method 29 Jo and Kims method 29 Jo and Kims method 29 Our method 12 Our method 12 Our method 12
MethodsImages Secrets PSNR Bit Rate Secrets PSNR Bit Rate
Lena 14905 28.39 0.5 15735 28.67 0.44
Boat 13893 27.74 0.5 15649 28.38 0.43
Jet 14288 28.81 0.5 15676 29.61 0.41
Pepper 15023 28.72 0.5 15726 29.37 0.42
Baboon 12162 23.54 0.5 13891 23.77 0.55
Sailboat 13863 27.50 0.5 15756 27.76 0.45
Codebook size 256State codebook size 16
SMVQ 9253VQ 6473No secrets 403
21

• Experiments

22
Data Hiding in VQ Codes

• Not every image block has the same capacity.

MELG(mean value)
PNNE(Euclidean distance)
ACE 20 (Cartesian product)
Codebook
23

• ACE method (Du and Hsu, 2003)

Secret data (001 1110)2 (30)10
Clustering result
5 3 8 6 3
Modified index table
24
Adaptive Embedding (1)

• Data reuse 13 To avoid the codeword waste in
  a group.

s2
s3
s1
s2
s1
6 1 8 5 2
Index table
00
0
00
5
6
7
1
2
3
4
5
6
7
1
2
3
4
8
0
01
1
01
1
10
0
10
0
11
11
Secret data (0 01 1 11)2
Clustering result
7 2 8 6 4
Modified index table
25

• Experiments

TH The number of clustering groups Lena Lena Jet Jet Pepper Pepper
TH The number of clustering groups Capacity (bit) PSNR (dB) Capacity (bit) PSNR (dB) Capacity (bit) PSNR (dB)
9 475 5926 32.16 11422 31.446 4689 31.339
12 449 10286 32.028 13014 31.382 9046 31.225
20 389 19658 31.532 19843 31.133 18623 30.798
35 287 32212 30.285 37244 29.869 29178 29.786
60 154 51444 27.664 45815 27.785 47339 27.409
90 73 65623 25.153 67842 25.258 66944 24.541
130 28 78935 21.798 73487 23.125 76689 21.894
Codebook size 512
26

• Experiments

27

• Experiments

VQ compressed image Payload(kb) Average payload (bit/block) PSNR (dB) PSNR (dB) PSNR (dB) PSNR (dB) PSNR (dB)
VQ compressed image Payload(kb) Average payload (bit/block) MGLE PNNE ACE Our method Our method
Lena(PSNR 32.243 dB) 4 0.25 28.834 31.053 31.487 32.185 32.189
Lena(PSNR 32.243 dB) 8 0.5 26.866 30.178 30.492 32.039 32.078
Lena(PSNR 32.243 dB) 16 1 24.555 28.778 29.817 31.683 31.762
Lena(PSNR 32.243 dB) 32 2 - - 28.588 29.942 30.311
Lena(PSNR 32.243 dB) 48 3 - - 27.613 26.994 27.867
Lena(PSNR 32.243 dB) 64 4 - - 26.401 24.220 25.153
Lena(PSNR 32.243 dB) 80 5 - - 24.610 18.641 21.968
Codebook size 512
Non-reuse
reuse
28

• Experiments

PSNR results at different embedding capacities
for the Lena image
29

• Experiments

Local results in the Lena image produced by
different hiding methods (capacity 16 kilobit)
30
Adaptive Embedding (2)

• Codeword movement 17 To increase payload
  capacity

31
Adaptive Embedding (2)

• Adaptive alternatives 17 To hide the secret
  bits in SMVQ codes

32

• Experiments

Utility rate of codewords in the sorted state
codebook by SMVQ
33

• Experiments

Codebook 512State codebook 16
34

• Experiments

PSNR results at different embedding capacities
for the Lena image
35
Reversible Data Hiding
Compressed codes 10111011 11..
11010110 01..
Information
Internet
Sender
Original codes 10111011 11..
Information
Receiver
36
Reversible Data Hiding

• Clustering of codeword-trios

Embeddable
indicator


37

• Two-bit extendable embedding

38

• Experiments

Codebook size 256
39

• Experiments

Methods Images VQ(0.5 bpp) Jo and Kims method Jo and Kims method Our method(1 bit / index) Our method(1 bit / index) Our method(2 bits / index) Our method(2 bits / index)
Methods Images PSNR PNSR Payload Payload Rate Payload Rate
Lena 31.338 28.514 15650 12016 0.54 23884 0.67
Jet 30.517 28.270 15388 13163 0.53 26170 0.67
Toys 29.862 26.520 15344 14316 0.51 28606 0.68
Pepper 30.712 27.758 15803 13248 0.53 26284 0.67
GoldHill 28.803 26.694 15656 9803 0.57 19420 0.68
Zelda 32.869 28.826 16115 12012 0.53 23754 0.67
Codebook size 256
40

• Experiments

41
Future Research Directions

• Fast VQ codebook search
• Other projection Voronoi diagram full-search
  equivalent
• SMVQ efficiency
• Reversible data hiding
• Construct a unique relation of one-to-one mapping
  
• Apply to other codes (SOC, STC, )
• Other VQ applications

42
Thanks for your attention
43

• HOSM scheme (Wang and Yang 56, 2005)
• Hierarchy-Oriented Searching Method
• Use the iterated-clustering concept to put the
  hierarchical structure together in order to
  create representative virtual codewords (non-leaf
  nodes) in a Tree structure.

44

• SCS scheme (Tai et al. 51, 1996)

45

• Shie et al.s scheme (2006)

Block diagram of the embedding procedure in 46
46

• GoldHill

47
Conclusions

• Part I
• Wavelet-Based Initialization for VQ Codebook
  Generation
• Fast Ripple Search Algorithm for VQ Codebook
• Part II
• Data Hiding Scheme Based on SMVQ
• Adaptive Embedding Scheme Based on VQ
• Reversible Hiding Scheme Based on VQ

48
VQ

• Overview

49
LBG Clustering Algorithm
Linde-Buzo-Gray iterative clustering
50
Wavelet-Based Initialization

• DWT the property of signal decomposition

One round 272752824
One round 272752822
51

• Experiments

MethodsCodebook size MethodsCodebook size MethodsCodebook size LBG method Our method Our method
MethodsCodebook size MethodsCodebook size MethodsCodebook size LBG method Phase 1 Phase 2
Set-I 256 Iteration 27 3 17
Set-I 256 PSNR 28.594 28.633 28.633
Set-I 512 Iteration 23 2 20
Set-I 512 PSNR 29.293 29.360 29.360
Set-II 256 Iteration 26 3 22
Set-II 256 PSNR 27.569 27.667 27.667
Set-II 512 Iteration 22 2 19
Set-II 512 PSNR 28.311 28.442 28.442