1. Reversible data hiding scheme based on significant-bit-difference expansion
Sourse: IET Image Processing ( Volume: 11, Issue: 11, ), Pages 1002 – 1014
Authors: Weiqing Wang, Junyong Ye , Tongqing Wang, Weifu Wang
Speaker: Guan-Long Li
Date: /12/7

2. Outline Related Work Proposed method Experiment Results Reference
Robust lossless image data hiding scheme
Proposed method
Experiment Results
Reference
Conclusion

3. Related Works Robust lossless image data hiding scheme
𝑀 𝑖,𝑗 = 𝑚𝑜𝑑 𝑖,2 =𝑚𝑜𝑑(𝑗,2) −1 𝑚𝑜𝑑(𝑖,2)≠𝑚𝑜𝑑(𝑗,2)
𝑎 (𝑘) = 𝑖=1 𝑚 𝑗=1 𝑛 ( 𝑐 𝑘 𝑖,𝑗 ×𝑀(𝑖,𝑗))

4. Proposed method Secrete data 𝐼 𝐻𝑆𝐵 𝐼′ 𝐻𝑆𝐵 𝐼 𝐿𝑆𝐵 Stego Image
Original Image

5. Proposed method n=3 HSB LSB a7 a6 a5 a4 a3 a2 a1 a0 a7 a6 a5 a4 a3 a2 a7 a6 a5 a4 a3 a2 a1 a0 a7 a6 a5 a4 a3 a2 a1 a0
120
n=3 15 1 1
𝑚𝑎𝑥=𝑓𝑙𝑜𝑜𝑟( 𝑛 )
𝑚𝑖𝑛=𝑓𝑙𝑜𝑜𝑟 𝑛 =0

6. 𝐼 𝐿𝑆𝐵 𝑖,𝑗 =𝐼 𝑖,𝑗 − 𝐼 𝐻𝑆𝐵 (𝑖,𝑗)× 2 𝑛
Proposed method 15 14
HSB
120
121
119
122
𝐼 𝐻𝑆𝐵 𝑖,𝑗 =𝑓𝑙𝑜𝑜𝑟 𝐼 𝑖,𝑗 2 𝑛
𝐼 𝐿𝑆𝐵 𝑖,𝑗 =𝐼 𝑖,𝑗 − 𝐼 𝐻𝑆𝐵 (𝑖,𝑗)× 2 𝑛 1 7 2
LSB

7. Proposed method 15 14 15 -1
𝑑 𝑖,𝑗 = 𝑑 𝑖,1 = 𝐼 𝐻𝑆𝐵 𝑖,1 − 𝐼 𝐻𝑆𝐵 𝑖−1,1 ,𝑖𝑓 𝑗=1 𝑑 𝑖,𝑗 = 𝐼 𝐻𝑆𝐵 𝑖,𝑗 − 𝐼 𝐻𝑆𝐵 𝑖,𝑗−1 ,𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

8. Proposed method Difference-Value intensity Expansion bin a: Maximum
b:Second Maximum
Embedd bit
Expansion bin
Shifted bin
Unused bin

9. Proposed method
𝐼 ′ 𝐻𝑆𝐵 𝑖,𝑗 = 𝐼 𝐻𝑆𝐵 𝑖,𝑗−1 +𝑑 𝑖,𝑗 +𝑡, 𝑖𝑓 𝑑 𝑖,𝑗 =𝑏 𝑎𝑛𝑑 𝐼 𝐻𝑆𝐵 𝑖,𝑗−1 +𝑑 𝑖,𝑗 <𝑚𝑎𝑥−1 𝐼 𝐻𝑆𝐵 𝑖,𝑗−1 +𝑑 𝑖,𝑗 −𝑡, 𝑖𝑓 𝑑 𝑖,𝑗 =𝑎 𝑎𝑛𝑑 𝐼 𝐻𝑆𝐵 𝑖,𝑗−1 +𝑑 𝑖,𝑗 > 𝐼 𝐻𝑆𝐵 𝑖,𝑗−1 +𝑑 𝑖,𝑗 +1, 𝑖𝑓 𝑑 𝑖,𝑗 >𝑏 𝑎𝑛𝑑 𝐼 𝐻𝑆𝐵 𝑖,𝑗−1 +𝑑 𝑖,𝑗 <𝑚𝑎𝑥−1 𝐼 𝐻𝑆𝐵 𝑖,𝑗−1 +𝑑 𝑖,𝑗 −1, 𝑖𝑓 𝑑 𝑖,𝑗 <𝑎 𝑎𝑛𝑑 𝐼 𝐻𝑆𝐵 𝑖,𝑗−1 +𝑑 𝑖,𝑗 > 𝐼 𝐻𝑆𝐵 𝑖,𝑗−1 +𝑑 𝑖,𝑗 , 𝑜𝑡ℎ𝑒𝑟𝑠
𝐼 ′ 𝐻𝑆𝐵 𝑖,1 = 𝐼 𝐻𝑆𝐵 𝑖−1,1 +𝑑 𝑖,1 +𝑡, 𝑖𝑓 𝑑 𝑖,𝑗 =𝑏 𝑎𝑛𝑑 𝐼 𝐻𝑆𝐵 𝑖−1,1 +𝑑 𝑖,1 <𝑚𝑎𝑥−1 𝐼 𝐻𝑆𝐵 𝑖−1,1 +𝑑 𝑖,1 −𝑡 , 𝑖𝑓 𝑑 𝑖,𝑗 =𝑎 𝑎𝑛𝑑 𝐼 𝐻𝑆𝐵 𝑖−1,1 +𝑑 𝑖,1 > 𝐼 𝐻𝑆𝐵 𝑖−1,1 +𝑑 𝑖,1 +1, 𝑖𝑓 𝑑 𝑖,𝑗 >𝑏 𝑎𝑛𝑑 𝐼 𝐻𝑆𝐵 𝑖−1,1 +𝑑 𝑖,1 <𝑚𝑎𝑥−1 𝐼 𝐻𝑆𝐵 𝑖−1,1 +𝑑 𝑖,1 −1, 𝑖𝑓 𝑑 𝑖,𝑗 <𝑎 𝑎𝑛𝑑 𝐼 𝐻𝑆𝐵 𝑖−1,1 +𝑑 𝑖,1 > 𝐼 𝐻𝑆𝐵 𝑖−1,1 +𝑑 𝑖, , 𝑜𝑡ℎ𝑒𝑟𝑠

10. Proposed method 15 14 15 -1 15 14 13 HSB with secret bits 1 a=0 b=1-1 15 14 13
HSB with secret bits 1
a=0 b=1
Secret bits

11. Proposed method 15 14 13
120
113
111
122
𝐼 ′ 𝑖,𝑗 = 𝐼 ′ 𝐻𝑆𝐵 𝑖,𝑗 × 2 𝑛 + 𝐼 𝐿𝑆𝐵 (𝑖,𝑗) 1 7 2

12. Proposed method 𝑑 ′ 𝑖,1 = 𝐼 ′ 𝐻𝑆𝐵 𝑖,1 − 𝐼 ′ 𝐻𝑆𝐵 𝑖−1,1
𝑑 ′ 𝑖,1 = 𝐼 ′ 𝐻𝑆𝐵 𝑖,1 − 𝐼 ′ 𝐻𝑆𝐵 𝑖−1,1
𝑑 ′ 𝑖,𝑗 = 𝐼 ′ 𝐻𝑆𝐵 𝑖,𝑗 − 𝐼 ′ 𝐻𝑆𝐵 𝑖,𝑗−1
𝐼 𝐻𝑆𝐵 𝑖,1 = 𝐼′ 𝐻𝑆𝐵 𝑖−1,𝑗 + 𝑑 ′ 𝑖,1 −1, 𝑖𝑓 𝑑 ′ 𝑖,1 >𝑏+1 𝑎𝑛𝑑 𝐼 ′ 𝐻𝑆𝐵 𝑖−1,1 +𝑑′ 𝑖,1 <255 𝐼′ 𝐻𝑆𝐵 𝑖−1,𝑗 + 𝑑 ′ 𝑖,1 +1, 𝑖𝑓 𝑑 ′ 𝑖,1 <𝑎 𝑎𝑛𝑑 𝐼 ′ 𝐻𝑆𝐵 𝑖−1,1 + 𝑑 ′ 𝑖,1 > 𝐼′ 𝐻𝑆𝐵 𝑖−1,𝑗 + 𝑑 ′ 𝑖, ,𝑜𝑡ℎ𝑒𝑟𝑠
𝐼 𝐻𝑆𝐵 𝑖,𝑗 = 𝐼′ 𝐻𝑆𝐵 𝑖,𝑗−1 + 𝑑 ′ 𝑖,𝑗 −1, 𝑖𝑓 𝑑 ′ 𝑖,𝑗 >𝑏+1 𝑎𝑛𝑑 𝐼 ′ 𝐻𝑆𝐵 𝑖,𝑗−1 +𝑑′ 𝑖,𝑗 <255 𝐼′ 𝐻𝑆𝐵 𝑖,𝑗−1 + 𝑑 ′ 𝑖,𝑗 +1, 𝑖𝑓 𝑑 ′ 𝑖,𝑗 <𝑎 𝑎𝑛𝑑 𝐼 ′ 𝐻𝑆𝐵 𝑖,𝑗−1 + 𝑑 ′ 𝑖,𝑗 > 𝐼′ 𝐻𝑆𝐵 𝑖,𝑗−1 + 𝑑 ′ 𝑖,𝑗 ,𝑜𝑡ℎ𝑒𝑟𝑠

13. Proposed method 15 14 13 15 14
120
121
119
122
120
113
111
122
𝑰′ 𝑯𝑺𝑩
𝑰 𝑯𝑺𝑩 15 -1 1 𝑰′ 𝑰 1 7 2
𝒅 ′ (𝒊,𝒋)
Secret bits
𝑰 𝑳𝑺𝑩

14. Experiment Results Various EC values Image n=1 n=2 n=3 n=4 01 7.9 8
8.8
7.5 02
5.2
5.4
6.3 5 03
6.6
6.7
6.4 04 6
7.2
5.9 05
3.5
3.8
5.1 3 06
3.9
3.6 07
4.3
4.1 08
3.7
4.6
3.4 09
6.5
7.4 10
4.9
5.8
The unit of EC is 10,000 bits

15. Experiment Results
Performance evaluation in terms of capacity–distortion curve

16. Experiment Results Corresponding PSNR values (dB) Capacity(bits)
10,000
20,000
30,000
40,000
Lena
1-PEE 60
56.8
54.6
53.3
proposed
61.6
58.2
55.5
53.7
Baboon
55.9
50.6 －
52.1
Airplane
63.8
60.4
58.3
63.6 61
59.3
57.3
Barbara
59.9
56.7
54.7
53.2
61.58
58.5
55.8
53.8
Elaine
50.5
61.2
56.4
54.9
Lake
58.7
54.2
51.3 54

17. References
[13] Yi, S., Zhou, Y.: ‘Binary-block embedding for reversible data hiding in encrypted images’, Signal Process., 2017, 133, pp. 40–51
[25] Li, X., Li, B., Yang, B., et al.: ‘General framework to histogram-shiftingbased reversible data hiding’, IEEE Signal Process. Lett., 2013, 22, (6), pp. 2181–2191
[34] Ma, X., Pan, Z., Hu, S., et al.: ‘High-fidelity reversible data hiding scheme based on multi-predictor sorting and selecting mechanism’, J. Vis. Commun. Image Represent., 2015, 28, pp. 71–82
[42] Wang, R.-Z., Lin, C.-F., Lin, J.-C.: ‘Hiding data in images by optimal moderately significant-bit replacement’, IEE Electron. Lett., 2000, 36, (25), pp. 2069–2070
[43] Zeng, X.T.: ‘A lossless robust data hiding scheme’, Pattern Recognit., 2010, 43, pp –1667

18. Conclusions
It exploits the higher pixel-bit-planes with importance within a local neighborhood to construct a more correlated pixel pair sequence, and addresses the question of how to determine the pixels into pairs properly for pairwise SBDE
Actually, this paper have the correct concept, but the results are wrong. In the data embedding, I get the different results when I follow the steps.