THE JOURNAL OF CHINA UNIVERSITIES OF POSTS AND TELECOMMUNICATIONS

Presentation on theme: "THE JOURNAL OF CHINA UNIVERSITIES OF POSTS AND TELECOMMUNICATIONS"— Presentation transcript:

1 THE JOURNAL OF CHINA UNIVERSITIES OF POSTS AND TELECOMMUNICATIONS
Rotation, scaling and translation invariant image watermarking using feature points THE JOURNAL OF CHINA UNIVERSITIES OF POSTS AND TELECOMMUNICATIONS Volume 15, Issue 2, June 2008 Reporter : 黃烱育 碩研資工一甲 M97G0217 黃烱育 2018/8/23

2 Author LI Lei-da(李雷達), Ph. D. Candidate,
Xidian University, interested in pattern recognition, information security and digital watermarking. GUO Bao-long(郭寶龍), Ph. D. and professor, Xidian University, interested in neural networks, pattern recognition, image processing and image communication. 碩研資工一甲 M97G0217 黃烱育 2018/8/23

3 Outline Introduction Feature-based invariant regions
Watermark insertion Watermark extraction Simulation results and discussions Conclusions 碩研資工一甲 M97G0217 黃烱育 2018/8/23

4 Introduction To protect the copyright of digital contents.
Robust to RST attacks Harris feature points and Image normalized Embedded into the invariant regions in discrete cosine transform (DCT) domain. 碩研資工一甲 M97G0217 黃烱育 2018/8/23

5 Feature-based invariant regions(1/6)
Harris corner detector 碩研資工一甲 M97G0217 黃烱育 2018/8/23

6 Feature-based invariant regions(2/6)
Invariant region generation 碩研資工一甲 M97G0217 黃烱育 2018/8/23

7 Feature-based invariant regions(3/6)
Invariant region generation 碩研資工一甲 M97G0217 黃烱育 2018/8/23

8 Feature-based invariant regions(4/6)
Invariant region generation 碩研資工一甲 M97G0217 黃烱育 2018/8/23

9 Feature-based invariant regions(5/6)
Rotation normalization 碩研資工一甲 M97G0217 黃烱育 2018/8/23

10 Feature-based invariant regions(6/6)
Modification for watermarking Transform into DCT domain Embedded into the modified patches. 碩研資工一甲 M97G0217 黃烱育 2018/8/23

11 Watermark insertion (1/2)
碩研資工一甲 M97G0217 黃烱育 2018/8/23

12 Watermark insertion (2/2)
Watermark bits Wi={w1,w2…,wn} Secret key K T: control watermark invisibility 碩研資工一甲 M97G0217 黃烱育 2018/8/23

13 Watermark extraction 碩研資工一甲 M97G0217 黃烱育 2018/8/23

14 Simulation results and discussions(1/2)
碩研資工一甲 M97G0217 黃烱育 2018/8/23

15 Simulation results and discussions(1/2)
碩研資工一甲 M97G0217 黃烱育 2018/8/23

16 Simulation results and discussions(1/2)
碩研資工一甲 M97G0217 黃烱育 2018/8/23

17 Conclusions A novel watermark synchronization scheme that combines the image LMSP (Locally Most Stable Points) and the image normalization is presented. The proposed scheme is robust to geometric attacks, signal processing attacks, as well as some combined attacks. 碩研資工一甲 M97G0217 黃烱育 2018/8/23

18 Thank you 碩研資工一甲 M97G0217 黃烱育 2018/8/23

19 Feature-based invariant regions(5/6)
碩研資工一甲 M97G0217 黃烱育 2018/8/23

20 Feature-based invariant regions(4/6)
Rotation normalization define two tensors t1 and t2 : Normalizing angle θ: Define another tensor t3 : 碩研資工一甲 M97G0217 黃烱育 2018/8/23

21 Feature-based invariant regions(3/6)
Scale normalization 碩研資工一甲 M97G0217 黃烱育 2018/8/23

22 Feature-based invariant regions(2/6)
A digital image f (x, y), its (p +q) th order moment is defined as ： The corresponding central moment is defined as： where is the centroid of the image. 碩研資工一甲 M97G0217 黃烱育 2018/8/23

23 Feature-based invariant regions
碩研資工一甲 M97G0217 黃烱育 2018/8/23