W. Puech (LIRMM) N. Hadj Said (SIMPA)

(1)

Authors: D. Goudia (LIRMM-SIMPA) M. Chaumont (LIRMM)

W. Puech (LIRMM)

N. Hadj Said (SIMPA)

(2)

• Few words about watermarking & compression

• The proposed joint JPEG2000 compression &

watermarking scheme

• Experimental evaluations

• Conclusions

D. Goudia A Joint JPEG2000 Compression & Watermarking System using a

TCQ-based Quantization Scheme

(3)

(4)

Message

(regarding work)

Watermark embedder

Watermark key

Original work

Watermark extractor Attacks

Extracted message

+

Watermark key

Watermarked work (looks like original)

Attacked watermarked work

A Joint JPEG2000 Compression & Watermarking System using a 1 TCQ-based Quantization Scheme

D. Goudia

(5)

Compressed/marked bitstream

Decompression Watermarking

decoder

Decompressed/marked image

Compression Watermarking

encoder Marked image

Extracted watermak

Degradation of the watermark & less robustness

(6)

Compressed/marked bitstream

Decompression Watermarking

decoder

Decompressed/marked image

Compression Watermarking

encoder Marked image

Extracted watermak

A Joint JPEG2000 Compression & Watermarking System using a 2 TCQ-based Quantization Scheme

D. Goudia

(7)

JOINT WATERMARKING

& COMPRESSION VISUAL QUALITY

HIGH PAYLOADS

LOW

COMPLEXITY

ROBUSTNESS

(8)

A Joint JPEG2000 Compression & Watermarking System using a TCQ-based Quantization Scheme

D. Goudia

(9)

The joint JPEG2000 encoder/watermark embedding scheme.

(10)

2 0

0

D D

A   A

₁

 D

₁

 D

₃

Partitioning of a scalar quantizer into 4 subsets conbined to form 2 union quantizers:

&

A Joint JPEG2000 Compression & Watermarking System using a 5 TCQ-based Quantization Scheme

D. Goudia

(11)

• Use of shifted TCQ quantizers

• 2 groups of union quantizers :

- Group 0 : The bit to be embedded is 0 - Group 1: The bit to be embedded is 1

• Modification of the trellis structure

0 3 0

1 0

1 0 2 0

0 0

0

D D , A D D

A    

1 3 1

1 1

1 1 2 1

0 1

0

D D , A D D

A    

The TCQ quantization used in the proposed joint

scheme

(12)

• Modification of the trellis structure

7 Labeling with the trellis

used in JPEG2000

Labeling with the modified

trellis used in our joint scheme

The TCQ quantization used in the proposed joint scheme

A Joint JPEG2000 Compression & Watermarking System using a TCQ-based Quantization Scheme

D. Goudia

(13)

• Modification of the trellis structure

(14)

The message is embedded by choosing between two TCQ quantizers from the 2 groups of union quantizers at each transition in the trellis

A Joint JPEG2000 Compression & Watermarking System using a 9 TCQ-based Quantization Scheme

D. Goudia

(15)

Quantization and watermark embedding

Step 1 Step 2 Step 3

Generation of the group 0 and group 1

shifted TCQ quantizers

Trellis pruning TCQ quantization

Viterbi Algorithm

(16)

Quantization and watermark embedding

Step 1 Step 2 Step 3

Generation of the group 0 and group 1

shifted TCQ quantizers

Trellis pruning TCQ quantization Viterbi Algorithm

A Joint JPEG2000 Compression & Watermarking System using a TCQ-based Quantization Scheme

D. Goudia

(17)

Trellis construction with branch labelling

Step 1

(18)

Quantization and watermark embedding

Step 1 Step 2 Step 3

Generation of the group 0 and group 1

shifted TCQ quantizers

Trellis pruning TCQ quantization Viterbi Algorithm

A Joint JPEG2000 Compression & Watermarking System using a TCQ-based Quantization Scheme

D. Goudia

(19)

Step 2 Trellis pruning

(20)

Quantization and watermark embedding

Step 1 Step 2 Step 3

Generation of the group 0 and group 1

shifted TCQ quantizers

Trellis pruning TCQ quantization Viterbi Algorithm

A Joint JPEG2000 Compression & Watermarking System using a TCQ-based Quantization Scheme

D. Goudia

(21)

Viterbi Algorithm

Step 3

(22)

• Dequantization by using the simplified trellis

During JPEG2000 decompression

• Discrete Wavelet Transform (DWT)

• Viterbi Algorithm applied onto the whole trellis

After JPEG2000 decompression

JPEG2000 bitstream

Decompressed/marked image

Watermark extraction

A Joint JPEG2000 Compression & Watermarking System using a 14 TCQ-based Quantization Scheme

D. Goudia

(23)

(24)

Evaluation protocol 1:

• 4 test images of size 512 x 512

• 5 levels of wavelet decomposition , one tile, no ROI coding

• Variation of the bitrate from 2.5 bpp to 0.2 bpp

• Payload = 1 bit (message) for 16 pixels

– 1024 bits embedded in the HL sub-band of the second resolution level

A Joint JPEG2000 Compression & Watermarking System using a 15 TCQ-based Quantization Scheme

D. Goudia

(25)

Bike image Lena image

Visual quality performances under various compression bitrates

(26)

Evaluation protocol 2:

• 200 images of size 512 x 512

• Robustness of the watermark: 4 attacks

– Gaussian filtering – Gaussian noise

– Valumetric scaling – JPEG attack

A Joint JPEG2000 Compression & Watermarking System using a 17 TCQ-based Quantization Scheme

D. Goudia

(27)

Gaussian filtering attack Gaussian noise attack

Attacks (1)

(28)

Valumetric scaling attack JPEG attack Attacks (2)

A Joint JPEG2000 Compression & Watermarking System using a 19 TCQ-based Quantization Scheme

D. Goudia

(29)

(30)

• Quantization & watermarking at the same time

• 2 types of watermark extraction

• Good compression performances

• Robust to JPEG2000 compression at low bitrates

• Robustness to attacks: depends on the value of Δ _TCQ

• Prospects:

– Sensitivity to valumetric & jpeg attacks – Integration of Turbo TCQ

– Consider ROI processing and progressive transmission functionnality

A Joint JPEG2000 Compression & Watermarking System using a 20 TCQ-based Quantization Scheme

D. Goudia

(31)

(32)

SPIE 2011: VIPC

01/25/2011

(33)

Group 0 :

Group 1:

] , 0 [ ]) , 0 [ ] [ ( ]

ˆ [ i Q

₀

x i d i d i

x

D

j

 



] 2 , 0 [ ] , 1 [ , ] , 1 [ ]) , 1 [ ] [ ( ]

ˆ [ i  Q

₁

x i  d i  d i d i  d i   x

D

j

0 1 0

0 & A A

1 1 1

0 & A

A

(34)