Docstoc

Robust Image Watermarking Based on Multiband Wavelets and

Document Sample
Robust Image Watermarking Based on Multiband Wavelets and Powered By Docstoc
					Robust Image Watermarking Based on
 Multiband Wavelets and Empirical
        Mode Decomposition

        Authors: Ning Bi, Qiyu Sun, Daren Huang,
                 Zhihua Yang, Jiwu Huang
        Adviser: 李育強
        Speaker: 楊智雁
        Date:2010/01/12
        received July 29, 2006; revised April 24,2007;
        Accepted AUGUST 2007




             南台科技大學            資訊工程系
Outline
     1    Introduction

          Multiband Wavelet Transformation and
     2    Emoirical Mode Decomposition

     3    Watermark Embedding


     4    Optimization of the Parameters


     5    Experimental Results


     6    Conclusion




2
   1.Introductuon

 The watermarking system has been viewed as a
  possible solution to control unauthorized duplication
  and redistribution of those multimedia data

 Robustness, perceptually invisibility and security are
  the basic requirements




    3
   1.Introductuon (c.)

 We select the mean trend of each subimage in the
  multiband wavelet domain,instead of the subimage
  itself, to embed the watermark bits




    4
   2.Multiband Wavelet Transformation And
     Empirical Mode Decomposition
 1. Multiband Discrete Wavelet Decomposition

                                              in
 scaling filter    H 0 ( )  ng 0 (n)e

                                              in
 wavelet filters   H1 ( )  ngl (n)e




     5
   2.Multiband Wavelet Transformation And
     Empirical Mode Decomposition (c.)
 In this paper, we will use the following parameterized
  multiband scaling and wavelet filters, where M is the
  dilation and λ is the parameter




    6
2.Multiband Wavelet Transformation And
  Empirical Mode Decomposition (c.)




7
   2.Multiband Wavelet Transformation And
     Empirical Mode Decomposition (c.)
 2. Empirical Mode Decomposition

 In the whole data set, the number of extrema and the
  number of zero crossing must either equal or differ at
  most by one

 At any point, the mean value of envelope defined by
  the local maxima and the envelope defined by the
  local minima is zero


    8
2.Multiband Wavelet Transformation And
  Empirical Mode Decomposition (c.)




9
   2.Multiband Wavelet Transformation And
     Empirical Mode Decomposition (c.)
 The EMD extracts the finest scale or the shortest
  period component from the signal step by step

 Our simulation shows that the mean trend is extremely
  stable for Gaussian noise and JPEG compression
  attack




    10
   3.Watermark Embedding

1. Multiband Discrete Wavelet Decomposition




2. Watermark Embedding Domain




    11
    3.Watermark Embedding (c.)
 L include an approximation of the original image, and
     F


  then embedding watermarks in those subimages may
  easily result in visual block effects

        considered as components with highest
    H F is

    frequency, and then the watermark may not be
    detected

 We select subimages in the subband M as our
                                         F

  favorable blocks to embed watermark bits in our
  watermarking scheme

         12
3.Watermark Embedding (c.)



     I   *
         A(i )    I A(i )  rA(i )  r
                                      *
                                      A(i )




13
   4. Optimization of the Paramerers

 We define the percentage of energy with middle and
  high frequency
                                  
                                 EH ( I )
                PI , M   ( ) :
                                 E(I )
 The lesser energy of those subimages the lesser
  influence of the watermark process to the image

 This also implies that larger watermark strength S can
  be added


    14
   4. Optimization of the Paramerers (c.)

 We observe that the watermarking strength S(P,I)
  decreases when the parameter P increases




    15
   5. Experimental Resuits

 JPEG Lossy Compression     Salt and pepper noise




 Gaussian Noise             Median Filtering

 Gaussian Noise



   16
   5. Experimental Resuits (c.)

 ConvFilter Attack




    17
   6. Conciusion

 The proposed scheme is robust against JPEG
  compression, Gaussian noise, salt and pepper noise,
  median filtering, and ConvFilter (Gaussian filtering
  and Sharpening)

 The proposed scheme has high BER percentage under
  some geometric distortion attacks




    18
6. Conciusion (c.)




19
南台科技大學   資訊工程系

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:3
posted:12/10/2011
language:
pages:20