1. LOGO www.themegallery.com Region of Interest Fragile Watermarking for Image Authentication 指導老師：陳炳彰老師 學生：林青照

2. www.themegallery.com 參考出處  題目： ” Region of Interest Fragile Watermarking for Image Authentication”  作者： Yiping Chu, Yin Zhang, Sanyuan Zhang and Xiuzi Ye  學校： College of Computer Science, State Key Lab of CAD&CG, Zhejiang University,  來自： 2006 IEEE

3. www.themegallery.com Abstract 1.Digital watermarking plays an important role in image verification and copyright protection. 2. A watermarking scheme for authenticating ROI (region of interest) of image has been proposed in the paper.

4. www.themegallery.com 1. Introduction The watermarking schemes as mentioned above have not reported to be able to watermark for local image authentication and copyright protection

5. www.themegallery.com 2. Related work  2.1. Fragile watermarking Schemes 1.Kundur et al. [14] present a fragile watermarking technique which embeds a watermark in the discrete wavelet domain by quantizing the corresponding coefficients.  tamper detection in localized spatial and frequency regions 2. Wu et al. present a secure semi-fragile watermarking for image authentication  based on integer wavelet transform 3. Ding et al propose a semi-fragile watermarking scheme, which is a wavelet-based chaotic semi-fragile watermarking scheme with more JPEG compression tolerance.

6. www.themegallery.com 2. Related work  2.2. Foreground extraction 1.The GVF snake  it has two significant advantages over the traditional snake formulation. 2. Sun et al.  present a matting approach to obtain foreground from complex background by solving Poisson PDE equations. Poisson matting and GrabCut have their respective advantages over each other. In the paper, we adopt Poisson matting to extract foreground.

7. www.themegallery.com 3. Proposed scheme Fig1.(a) Diagram of watermark embedding scheme

8. www.themegallery.com 3. Proposed scheme (b) Diagram of watermark extraction and tamper detection scheme

9. www.themegallery.com 3. Proposed scheme  3.1. Reference mask generation  In order to produce the binary reference mask image, copyright owner adopt Poisson matting [18] to obtain the image RIM which consists of the ROI of original image as foreground and blue color (or other colors) background. Fig. 2. Reference mask image generation for Lena (a) ROI image obtain by using passion matting; (b) Binary image converted from (a); (c) Morphology filtering result of (b)

10. www.themegallery.com 3. Proposed scheme  3.2. Watermark embedding  Kundur et al. Firstly, we define a quantization function for our watermark embedding function. Quantization function  Our watermark embedding function is defined as follows

11. www.themegallery.com 3. Proposed scheme  3.3. Watermark detection  watermark detection, copyright owner use binary reference mask image BM that was preserve in the step of watermark embedding and the same parameter K as random seed to retrieve binary image watermark ' W.

12. www.themegallery.com 4. Experiments and results

13. www.themegallery.com 4. Experiments and results Fig. 4. (a) Tamper inside ROI ; (b) Embedding watermark ' W ; (c)Extracted watermark " W ; (d) Tamper location inside ROI revealed by T

14. www.themegallery.com 4. Experiments and results Figure 5. (a) Tamper in the exterior of ROI; (b) Embedding watermark ' W ; (c) Extracted watermark " W ; (d) T =0, no alarm occurred;

15. www.themegallery.com 5. Conclusion  A secure watermarking for ROI of image authentication based on foreground extraction has been presented.  Experiment results have demonstrated that the proposed scheme is capable of detecting tamper in the interior of ROI and tolerating tamper exterior of ROI.

16. LOGO www.themegallery.com Add your company slogan