C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and...

15
1 C M S 2001 Darmstad Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole Nationale Supérieure des Télécommunications 46, rue Barrault, 75634 Paris Cedex 13, France {zinger, jin, maitre}@tsi.enst.fr * Boğaziçi University 80815 Bebek - Istanbul, Turkey [email protected] S.Zinger, Z.Jin, B.Sankur*, H. Maître

Transcript of C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and...

Page 1: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

1

C M S 2001 Darmstadt

Optimization of Watermarking Performances using Error

Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître

Ecole Nationale Supérieure des Télécommunications

46, rue Barrault, 75634 Paris Cedex 13, France

{zinger, jin, maitre}@tsi.enst.fr

* Boğaziçi University

80815 Bebek - Istanbul, Turkey

[email protected]

S.Zinger, Z.Jin, B.Sankur*, H. Maître

Page 2: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

2

C M S 2001 Darmstadt

The goal

to explore applying coding methods

for watermarking purposes

robustnessthe length of the

signaturethe length of the

coded word

Page 3: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

3

C M S 2001 Darmstadt

The constraints:

• imperceptibility

• robustness

• capacityw ~ 10 - 100 bits

c ~ 200 - 500 bits

Page 4: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

4

C M S 2001 Darmstadt

Capacity is the maximum amount of bits which can be hidden in an image

• Ramkumar, 1998• Servetto, 1998• Barni, 1999• Marvel, 1999• Lin, 2001• Sugihara, 2001

Page 5: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

5

C M S 2001 Darmstadt

Coding methods• Repetition• ECC BCH codes• Hybrid coding

original watermark

binary symmetric channel

extracted watermark

?

attacks

Page 6: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

6

C M S 2001 Darmstadt

Repetition

Bit error probability after repetitions r

,11

2

r

ri

irbsc

ibsc

irrep ppCP

wreprepsig PP 11,

Probability of error for a signature of bitsw

cwr

Page 7: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

7

C M S 2001 Darmstadt

BCH codes

• Standard BCH codes ,

• BCH codes by parts

• BCH codes with subtraction

n

ti

inbsc

ibsc

inBCHsig ppCP

1, )1(

,cn tkn ,, wk

Page 8: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

8

C M S 2001 Darmstadt

Examples of BCH codes

Capacity bits, payload bits500c 32w

Standard BCH codes

code (255,37,45)

BCH codes by parts

(127,15,27) (127,15,27) (127,15,27)

BCH codes with subtraction

(500,38,93)

Page 9: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

9

C M S 2001 Darmstadt

Hybrid coding

• BCH coding after repetition

• repetition after BCH

1

1, 1

n

repirep

n

ti

inhybridsig PPCP

Page 10: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

10

C M S 2001 Darmstadt

A comparison of performancesChannel error rate is 5%, capacity is 350 bits

210

3610Hybrid coding

Repetition

BCH with subtraction

Page 11: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

11

C M S 2001 Darmstadt

Average of performances 500200 c

Channel error rate is 5%

Hybrid coding

Repetition

BCH with subtraction

210

2510

for capacity

Page 12: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

12

C M S 2001 Darmstadt

Average of performances

500200 cChannel error rate is 15%

Hybrid coding

for capacity

Repetition

Page 13: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

13

C M S 2001 Darmstadt

Average of performances

500200 cChannel error rate is 30%

Hybrid coding

Repetition

for capacity

Page 14: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

14

C M S 2001 Darmstadt

The best coding strategy,10 2sigP 400c

Repetition

Shannon capacity

Hybrid coding BCH with

subtraction

for

Page 15: C M S 2001 Darmstadt 1 Optimization of Watermarking Performances using Error Correcting Codes and Repetition S.Zinger, Z.Jin, B.Sankur*, H. Maître Ecole.

15

C M S 2001 Darmstadt

Connection between channel error rate and JPEG compression rate

-0,1

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0 20 40 60

JPEG compression rate

be

r