Docstoc

Mean quantization based image watermarking

Document Sample
Mean quantization based image watermarking Powered By Docstoc
					 Mean quantization based
  image watermarking
Source: Image and Vision Computing, vol. 21,
        no. 8, Aug. 2003, pp. 717-727.
      Authors: L.-H. Chen and J.-J. Lin
           Speaker: Wei-Liang Tai




                                               1
             Outline

•   Introduction to Watermarking
•   Introduction to Wavelet Transforms
•   The Proposed Method
•   Experimental Results
•   Conclusions


                                         2
Introduction -
Watermarking
          Watermarking



    Invisible        Visible
  Watermarking     Watermarking




                                  3
      Introduction -
    Watermarking (Cont.)
• Invisible Watermarking
  – Robust Watermarking
     • It can against image processing such as filtering,
       lossy compression, sharpening, and so on.
  – Fragile Watermarking
     • It can detect any tiny alteration to the pixel value.
  – Semi-fragile Watermarking
     • It is moderately robust to lossy compression such as
       JPEG.



                                                               4
              Introduction –
            Wavelet Transforms
Haar Wavelet Transform
Phase 1) Horizontal :

    A   B   C   D       A+B   C+D   A-B   C-D      ㄅ ㄆ ㄇ ㄈ
    E   F   G   H       E+F   G+H   E-F   G-H      ㄉ ㄊ ㄋ ㄌ
    I   J   K   L       I+J   K+L   I-J   K-L      ㄍ ㄎ ㄏ ㄐ
    M   N   O   P       M+N   O+P   M-N   O-P
                                                   ㄑ ㄒ ㄓ ㄔ
Phase 2) Vertical :

                ㄅ ㄆ ㄇ ㄈ             ㄅ+ㄉ     ㄆ+ㄊ   ㄇ+ㄋ   ㄈ+ㄌ
                ㄉ ㄊ ㄋ ㄌ             ㄍ+ㄑ     ㄎ+ㄒ   ㄏ+ㄓ   ㄐ+ㄔ
                ㄍ ㄎ ㄏ ㄐ             ㄅ-ㄉ     ㄆ-ㄊ   ㄇ-ㄋ   ㄈ-ㄌ
                ㄑ ㄒ ㄓ ㄔ             ㄍ-ㄑ     ㄎ-ㄒ   ㄏ-ㄓ   ㄐ-ㄔ
                                                              5
     Introduction –
Wavelet Transforms (Cont.)




       JND (Just Noticeable Difference)
          Range = [ T-JND, T+JND ]
                                          6
                The Proposed Method
    Watermark W=w1w2…wL              Embed in subbands LL1 and HH1
    N where is the number of coefficients used to encode a watermark bit
      ex: N=3, w1=0                            Normalize, and compute mean value
       HH1                    JND                       4 6 7
                                                          
4
6     7
      6    9
           7    2     16   10   20    25               16 10 20  0.4
                                                           3
3     -2   0    4     18   25   26    16

3     6    3    6     18   10   18    10

0     3    -4   4     26   18   16    16



                    Extract
                                                         0.5  0.4  0.1
                     6 7     9
                                                    Inverse - normalize
                    16 10 20  0.5
                          3                    4               6     
                                                 0.1 16  6,   0.1 10  7
                     w1  0                    16              10   
                                                7    
                                                 0.1  20  9                    7
                                                20   
Experimental Results




  Original image   Watermarked image




                                       8
Experimental Results (Cont.)




                               9
           Conclusions

• Exploiting human visual system (HVS)
• Robust to common attacks




                                         10

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:8
posted:9/1/2012
language:English
pages:10