CN102682418B

CN102682418B - Method for embedding and extracting multiple zero watermarks of digital image

Info

Publication number: CN102682418B
Application number: CN201210144040.4A
Authority: CN
Inventors: 周亚训; 金炜; 陈芬; 徐星辰; 郑世超; 殷丹丹
Original assignee: Ningbo University
Current assignee: Ningbo University
Priority date: 2012-05-11
Filing date: 2012-05-11
Publication date: 2014-06-25
Anticipated expiration: 2032-05-11
Also published as: CN102682418A

Abstract

The invention discloses a method for embedding and extracting multiple zero watermarks of a digital image. Multiple zero watermarks are embedded and extracted in a discrete wavelet transform and discrete cosine transform composite domain. The multiple zero watermarks are embedded by constructing a binary watermark secret key by using a coefficient in an original digital image composite domain and then carrying out XOR operation with multiple actual binary digital watermarks, and the multiple zero watermarks are extracted by constructing a binary watermark private key by using a coefficient in a to-be-tested digital image composite domain and combining with zero watermark information. The method has the advantages of excellent watermark robustness and capability of resisting conventional image processing attack; and meanwhile, the method for embedding has the advantages that a plurality of binary digital watermarks are registered in a copyright database without any changing an original digital image, therefore, the embedded binary digital watermarks are not completely detected, the contradiction between the robustness and the difficulty in detection of the digital watermarks is balanced, and multiple requirements of application of the digital watermarks are met.

Description

A kind of multiple zero watermark embedding and extracting method of digital picture

Technical field

The present invention relates to a kind of zero digital watermark, especially relate to a kind of multiple zero watermark embedding and extracting method of digital picture.

Background technology

Digital watermarking, as a kind of technological means of effective solution Digital Media Works copyright protection and entitlement discriminating, is the study hotspot of Contemporary Digital media information security fields.Digital watermarking will be brought into play due effect, must possess two fundamentals of robustness and imperceptible.Watermark robustness, refers to that Digital Media is through after conventional digital signal processing or external attack, and the digital watermarking of embedding still has the aspect information such as copyright that good detectability still can reflect original figure media in other words.Watermark imperceptible, refers to that the embedding of digital watermarking can not have influence on the sense of hearing or the visual quality of original figure media, thereby can not affect the using value of original figure media.Obviously, traditional digital watermark technology, carry out certain modification by the spatial domain data to original figure media or coefficient in transform domain, certain characteristic information that comprises author or works is for example signed, copyright, sequence number, date or icon etc. are embedded in original figure media as digital watermarking, inevitably exist the contradiction between watermark robustness and watermark imperceptible: on the one hand, watermark robustness require in original figure media as much as possible embed digital watermark information with resist various conventional processing or have a mind to attack, and on the other hand, watermark imperceptible wishes that embed digital watermark information is as few as possible to avoid causing and the notable difference of original figure media, this is particularly important for some responsive digital pictures such as medical image and palmprint image etc., because the details pixel packets of this class digital picture has contained very important information, the distortion that in this class digital picture, the variation of any pixel causes all can affect the judgement to original digital image.

In recent years, the proposition of zero digital watermark had solved the contradiction between watermark robustness and the watermark imperceptible in traditional digital watermark technology well, became a new research branch in digital watermark technology.So-called zero digital watermark, construct a watermark keys by the unique characteristics of original figure media exactly, again to the to be embedded digital watermarking being of practical significance in conjunction with forming relevant zero watermark information post-registration to a kind of digital watermark technology in intellecture property watermark information database, it does not change to original figure media.Therefore, in zero digital watermark, the structure of zero watermark and registration process are exactly the digital watermark embed process on ordinary meaning, once register, original figure media are just considered to comprise real figure watermark and have possessed corresponding copyright protection ability.Because the real figure watermark in zero digital watermark is to be registered in intellecture property watermark information database, rather than be embedded in original figure media, also any problem that does not just exist original figure media quality to decline, just because of this characteristic, zero digital watermark well balance the contradiction between watermark robustness and the watermark imperceptible in traditional digital watermark technology.

At present, existing multiple zero watermarking project is suggested.2008, Zeng Fanjuan, Zhou Anmin have proposed a kind of digital picture zero watermarking project based on Contourlet conversion and svd on computer utility publication, first this scheme carries out Contourlet conversion to original image, be decomposed into the sub-band images of a series of multiple dimensioned, localization, directivity, select again low frequency sub-band to carry out piecemeal svd, construct zero watermark according to the unchangeability of the integer-bit size of first singular value in every decomposition.2009, the people such as Zhao Jie, king's Xuan and He Bing have proposed a kind of zero watermarking project based on image scrambling and wavelet transformation on computer engineering and science publication, first this scheme carries out scramble processing to watermarking images, then utilize low frequency subgraph after carrier image one-level wavelet decomposition and the watermarking images after scramble to obtain extracting matrix, carry out certain encryption that upsets to extracting matrix again, complete thus embedding and the extraction of zero watermark.2010, Bi Xiuli, He Chunxiang and Cheng Cheng have proposed a kind of zero watermarking project based on log-polar coordinate mapping and Lifting Wavelet on computer engineering and science publication, first this scheme carries out scramble by watermarking images, carrier image is carried out to convergent-divergent normalized, then carry out log-polar coordinate mapping, then carry out the lifting of one-level integer wavelet, after the low frequency subgraph after finally utilization is decomposed and scramble, watermarking images constructs zero watermark.2011, Wang Wujun has proposed zero watermarking project of a kind of multistage discrete cosine transform and svd on computing machine and digital engineering publication, first this scheme carries out multistage discrete cosine transform to the original image that will embed, extract low frequency subgraph and it is carried out to svd, secondly the watermarking images embedding is carried out to scramble, watermarking images after scramble is carried out to svd, finally the result after original image and twice svd of watermarking images is carried out to computing and produce zero watermark.

But zero watermarking project having proposed is at present the digital watermark technology that embeds and extract based on single zero watermark substantially, exist the limitation of function singleness.Embed multiple digital watermarkings to meet the needs of different application object owing to existing more in practical application in Digital Media works, such as width Digital Media works need to have two people or many people's signature with Shared Copyright after completing, or width Digital Media works are being issued, sell, the different phases such as use need to indicate that the different owners are (as creator, publisher, user etc.) use legal information, and for example the author of Digital Media works (or publisher) embeds for the robust watermarking of copyright authentication with for fragile watermark of content authentication etc. in works simultaneously.Therefore, consider from the application diversity of digital watermarking, the zero watermarking project research that embeds multiple digital watermarkings has more actual application value and wide application prospect.

Summary of the invention

Technical matters to be solved by this invention is to provide a kind of multiple zero watermark embedding and extracting method of digital picture, it has been realized and in original digital image, has embedded multiple zero watermarks, and multiple zero watermarks that embed have excellent robust performance, effectively guarantee that the quality of original digital image is unaffected simultaneously.

The present invention solves the problems of the technologies described above adopted technical scheme: a kind of multiple zero watermark embedding methods of digital picture, is characterized in that comprising the following steps:

1.-1, at multiple zero watermark built-in ends, the gray level image that the original digital image of supposing K to be embedded bi-level digital watermark is 8bit, and be designated as F, F={0≤f (m, n)≤255, 1≤m≤M, 1≤n≤N}, wherein, K >=2, M represents the vertical resolution of the original digital image F of K to be embedded bi-level digital watermark, N represents the lateral resolution of the original digital image F of K to be embedded bi-level digital watermark, M × N represents the resolution of the original digital image F of K to be embedded bi-level digital watermark, f (m, n) represent that in the original digital image F of K to be embedded bi-level digital watermark, coordinate position is (m, the pixel value of pixel n),

1. ,-2, at multiple zero watermark built-in ends, suppose that K to be embedded bi-level digital watermark is all bianry image, and be designated as W respectively ₁, W ₂..., W _k... W _k, for k bi-level digital watermark W to be embedded _k, W _k={ w _k(i _k, j _k)=0 or 1,1≤i _k≤ I _k, 1≤j _k≤ J _k, wherein, 1≤k≤K, I _krepresent k bi-level digital watermark W to be embedded _kvertical resolution, J _krepresent k bi-level digital watermark W to be embedded _klateral resolution, I _k× J _krepresent k bi-level digital watermark W to be embedded _kresolution, w _k(i _k, j _k) represent k bi-level digital watermark W to be embedded _kmiddle coordinate position is (i _k, j _k) the pixel value of pixel;

1.-3, the original digital image F of K to be embedded bi-level digital watermark is normalized, obtain the digital picture after normalized, be designated as F ', be (m by middle the digital picture F ' after normalized coordinate position, the pixel value of pixel n) is designated as f ' (m, n), f ' (m, n)=f (m, n)/255;

1.-4, F ' is carried out to L level two-dimensional discrete wavelet conversion, obtain a first small echo ll channel and multiple the first wavelet details subgraph, the first small echo ll channel is designated as to FA, wherein, the resolution of FA is (M/2 ^l) × (N/2 ^l),

min () is for getting minimum value function, and max () is for getting max function, symbol

represent to get the maximum integer that is less than himself, I ₁represent the 1st bi-level digital watermark W to be embedded ₁vertical resolution, J ₁represent the 1st bi-level digital watermark W to be embedded ₁lateral resolution, I _krepresent K bi-level digital watermark W to be embedded _kvertical resolution, J _krepresent K bi-level digital watermark W to be embedded _klateral resolution;

1.-5, FA is carried out to two-dimension discrete cosine transform, obtain one with the first Two Dimension Discrete Cosine matrix of FA equal resolution, be designated as FAC, then FAC being carried out to Zig-Zag scanning arranges, obtain a first one-dimensional discrete cosine transform coefficient sequence, be designated as FACS, FACS={facs (x), 1≤x≤(M/2 ^l) × (N/2 ^l), wherein, facs (x) represents x discrete cosine transform coefficient in FACS, the 2nd discrete cosine transform coefficient in FACS starts to be discrete cosine transform ac coefficient;

1. ,-6, successively according to the resolution of each bi-level digital watermark to be embedded, from FACS, select respectively and meet the I imposing a condition ₁× J ₁, I ₂× J ₂..., I _k× J _k... and I _k× J _kindividual discrete cosine transform ac coefficient, forms K the first one-dimensional discrete cosine transform ac coefficient sequence, and correspondence is designated as FACS respectively ₁, FACS ₂..., FACS _k... and FACS _k, for the satisfied I imposing a condition selecting from FACS _k× J _kthe k that individual discrete cosine transform ac coefficient forms a first one-dimensional discrete cosine transform ac coefficient sequence FACS _k, FACS _k={ facs _k(y), 1≤y≤I _k× J _k, then record FACS ₁, FACS ₂..., FACS _k... and FACS _kin the correspondence position information of various discrete cosine transform ac coefficient in FACS, wherein, I ₂represent the 2nd bi-level digital watermark W to be embedded ₂vertical resolution, J ₂represent the 2nd bi-level digital watermark W to be embedded ₂lateral resolution, facs _k(y) represent FACS _kin y discrete cosine transform ac coefficient, impose a condition and be more than or equal to the difference threshold of setting for the absolute value of difference of two the discrete cosine transform ac coefficients of arbitrary neighborhood in the first one-dimensional discrete cosine transform ac coefficient sequence that makes to form;

1.-7, compare respectively FACS ₁, FACS ₂..., FACS _k... and FACS _kin the size of two discrete cosine transform ac coefficients of arbitrary neighborhood, and return to logical value 1 or 0 according to comparative result, for FACS _kin two discrete cosine transform ac coefficients of arbitrary neighborhood, it is designated as respectively to facs _kand facs (z) _k(z+1), judge facs _k(z) >facs _k(z+1) whether set up, if set up, return to logical value 1, otherwise, return to logical value 0, wherein, 1≤z≤I _k× J _k-1; Then compare FACS ₁, FACS ₂..., FACS _k... and FACS _kin last discrete cosine transform ac coefficient and the size of the 1st discrete cosine transform ac coefficient, and return to logical value 1 or 0 according to comparative result, for FACS _kin last discrete cosine transform ac coefficient and the 1st discrete cosine transform ac coefficient, if the former is large, return to logical value 1, otherwise, return to logical value 0; Again according to FACS ₁, FACS ₂..., FACS _k... and FACS _keach self-corresponding logical value of returning, builds the first bi-level digital watermark keys one to one, for FACS _kthe corresponding logical value of returning, it is I that the logical value of returning is stored in to a size by Row Column order arrangement mode _k× J _ktwo-dimensional matrix in, the first bi-level digital watermark keys using this two-dimensional matrix as k, is designated as WB _k;

1.-8, to K to be embedded bi-level digital watermark W ₁, W ₂..., W _k..., W _kcarry out respectively scramble processing, K the bi-level digital watermark obtaining after scramble is processed respectively correspondence is designated as WS ₁, WS ₂..., WS _k... and WS _k, then by WS ₁, WS ₂..., WS _k... and WS _krespectively with K the first bi-level digital watermark keys WB ₁, WB ₂..., WB _k... and WB _kcorrespondence is carried out XOR and is obtained K zero watermark information one by one, and correspondence is designated as WO respectively ₁, WO ₂..., WO _k... and WO _k, WO ₁=xor (WS ₁, WB ₁), WO ₂=xor (WS ₂, WB ₂) ..., WO _k=xor (WS _k, WB _k) ..., WO _k=xor (WS _k, WB _k), then by K zero watermark information WO ₁, WO ₂..., WO _k... and WO _kbe stored in the digital watermark information database of registration body, complete K bi-level digital watermark W ₁, W ₂..., W _k..., W _kembedding, wherein, WS ₁represent W ₁carry out the bi-level digital watermark obtaining after scramble processing, WS ₂represent W ₂carry out the bi-level digital watermark obtaining after scramble processing, WS _krepresent W _kcarry out the bi-level digital watermark obtaining after scramble processing, WS _krepresent W _kcarry out the bi-level digital watermark obtaining after scramble processing, WB ₁represent the 1st the first bi-level digital watermark keys, WB ₂represent the 2nd the first bi-level digital watermark keys, WB _krepresent K the first bi-level digital watermark keys, xor () is XOR function;

1.-9, at multiple zero watermark built-in ends, by the FACS of record ₁, FACS ₂..., FACS _k... and FACS _kin the positional information of various discrete cosine transform ac coefficient in FACS, a K zero watermark information WO ₁, WO ₂..., WO _k..., WO _kand K bi-level digital watermark W ₁, W ₂..., W _k..., W _kbe transferred to multiple zero watermark extracting ends.

FACS in described step 1.-6 ₁, FACS ₂..., FACS _k... and FACS _kin the absolute value of difference of two discrete cosine transform ac coefficients of arbitrary neighborhood be more than or equal to respectively δ ₁, δ ₂..., δ _k... and δ _k, wherein, δ ₁represent for FACS ₁in two discrete cosine transform ac coefficients of arbitrary neighborhood difference absolute value set the 1st difference threshold, δ ₂represent for FACS ₂in two discrete cosine transform ac coefficients of arbitrary neighborhood difference absolute value set the 2nd difference threshold, δ _krepresent for FACS _kin two discrete cosine transform ac coefficients of arbitrary neighborhood difference absolute value set k difference threshold, δ _krepresent for FACS _kin two discrete cosine transform ac coefficients of arbitrary neighborhood difference absolute value set K difference threshold.

Multiple zero watermark extracting methods for digital picture, is characterized in that comprising the following steps:

2.-1, at multiple zero watermark extracting ends, the digital picture of K to be extracted bi-level digital watermark is designated as to TF, TF={0≤tf (m ', n ')≤255, 1≤m '≤M ', 1≤n '≤N ' }, wherein, K >=2, M ' represents the vertical resolution of the digital picture TF of K to be extracted bi-level digital watermark, N ' represents the lateral resolution of the digital picture TF of K to be extracted bi-level digital watermark, M ' × N ' represents the resolution of the digital picture TF of K to be extracted bi-level digital watermark, the resolution of the digital picture TF of K to be extracted bi-level digital watermark and the embedding of multiple zero watermark built-in ends have the resolution of digital picture of digital watermarking identical, tf (m ', n ') represent in the digital picture TF of K to be extracted bi-level digital watermark coordinate position for (m ', n ') the pixel value of pixel,

2. ,-2, at multiple zero watermark extracting ends, K to be extracted bi-level digital watermark is designated as respectively to W ' ₁, W ' ₂..., W ' _k... and W ' _k, for k bi-level digital watermark W ' to be extracted _k, W ' _k=w ' _k(i ' _k, j ' _k)=0 or 1,1≤i ' _k≤ I ' _k, 1≤j ' _k≤ J ' _k, wherein, 1≤k≤K, I ' _krepresent k bi-level digital watermark W ' to be extracted _kvertical resolution, J ' _krepresent k bi-level digital watermark W ' to be extracted _klateral resolution, I ' _k× J ' _krepresent k bi-level digital watermark W ' to be extracted _kresolution, k bi-level digital watermark W ' to be extracted _kresolution identical with the resolution of k bi-level digital watermark that multiple zero watermark built-in ends embed, w ' _k(i ' _k, j ' _k) represent k bi-level digital watermark W ' to be extracted _kmiddle coordinate position be (i ' _k, j ' _k) the pixel value of pixel;

2.-3, the digital picture TF of K to be extracted bi-level digital watermark is normalized, obtain the digital picture after normalized, be designated as TF ', by middle the digital picture TF ' after normalized coordinate position be (m ', n ') the pixel value of pixel be designated as tf ' (m ', n '), tf ' (m ', n ')=tf (m ', n ')/255;

2.-4, TF ' is carried out to L ' level two-dimensional discrete wavelet conversion, obtain a second small echo ll channel and multiple the second wavelet details subgraph, the second small echo ll channel is designated as to TFA, wherein, the resolution of TFA is (M '/2 ^l') × (N '/2 ^l'),

represent to get the maximum integer that is less than himself, I ₁' represent the 1st bi-level digital watermark W to be extracted ₁' vertical resolution, J ₁' represent the 1st bi-level digital watermark W to be extracted ₁' lateral resolution, I _k' represent K bi-level digital watermark W ' to be extracted _kvertical resolution, J _k' represent K bi-level digital watermark W ' to be extracted _klateral resolution;

2.-5, TFA is carried out to two-dimension discrete cosine transform, obtain one with the second Two Dimension Discrete Cosine matrix of TFA equal resolution, be designated as TFAC, then TFAC being carried out to Zig-Zag scanning arranges, obtain a second one-dimensional discrete cosine transform coefficient sequence, be designated as TFACS, TFACS={tfacs (x '), 1≤x '≤(M '/2 ^l') × (N '/2 ^l'), wherein, tfacs (x ') represents the individual discrete cosine transform coefficient of x ' in TFACS, the 2nd discrete cosine transform coefficient in TFACS starts to be discrete cosine transform ac coefficient;

2.-6, according to the FACS of multiple zero watermark built-in end records ₁, FACS ₂..., FACS _k... and FACS _kin the correspondence position information of various discrete cosine transform ac coefficient in FACS, from TFACS, extract respectively the I of correspondence position ₁' × J ₁', I ₂' × J ₂' ..., I _k' × J _k' ... and I _k' × J _k' individual discrete cosine transform ac coefficient, forms K the second one-dimensional discrete cosine transform ac coefficient sequence, and correspondence is designated as TFACS respectively ₁, TFACS ₂..., TFACS _k... and TFACS _k, for the FACS recording according to multiple zero watermark built-in ends _kin the correspondence position information of various discrete cosine transform ac coefficient in FACS, from TFACS, extract the I of correspondence position _k' × J _kk the second one-dimensional discrete cosine transform ac coefficient sequence TFACS of ' individual discrete cosine transform ac coefficient formation _k, TFACS _k={ tfacs _k(y '), 1≤y '≤I _k' × J _k', wherein, I _k' represent k bi-level digital watermark W ' to be extracted _kvertical resolution, J _k' represent k bi-level digital watermark W ' to be extracted _klateral resolution, tfacs _k(y ') represents TFACS _kin the individual discrete cosine transform ac coefficient of y ';

2.-7, compare respectively TFACS ₁, TFACS ₂..., TFACS _k... and TFACS _kin the size of two discrete cosine transform ac coefficients of arbitrary neighborhood, and return to logical value 1 or 0 according to comparative result, for TFACS _kin two discrete cosine transform ac coefficients of arbitrary neighborhood, it is designated as respectively to tfacs _k(z ') and tfacs _k(z '+1), judges tfacs _k(z ') >tfacs _kwhether set up (z '+1), if set up, returns to logical value 1, otherwise, return to logical value 0, wherein, 1≤z '≤I _k' × J _k'-1; Then compare TFACS ₁, TFACS ₂..., TFACS _k... and TFACS _kin last discrete cosine transform ac coefficient and the size of the 1st discrete cosine transform ac coefficient, and return to logical value 1 or 0 according to comparative result, for TFACS _kin last discrete cosine transform ac coefficient and the 1st discrete cosine transform ac coefficient, if the former is large, return to logical value 1, otherwise, return to logical value 0; Again according to TFACS ₁, TFACS ₂..., TFACS _k... and TFACS _keach self-corresponding logical value of returning, builds the second bi-level digital watermark keys one to one, for TFACS _kthe corresponding logical value of returning, it is I that the logical value of returning is stored in to a size by Row Column order arrangement mode _k' × J _k' two-dimensional matrix in, the second bi-level digital watermark keys using this two-dimensional matrix as k, is designated as TWB _k;

2.-8, by the K from multiple zero watermark built-in ends zero watermark information WO ₁, WO ₂..., WO _k... and WO _k, respectively with K the second bi-level digital watermark keys TWB ₁, TWB ₂..., TWB _k... and TWB _kcorrespondence is carried out XOR one by one, recovers to obtain K bi-level digital watermark, and correspondence is designated as TW respectively ₁, TW ₂..., TW _k... and TW _k, TW ₁=xor (WO ₁, TWB ₁), TW ₂=xor (WO ₂, TWB ₂) ..., TW _k=xor (WO _k, TWB _k) ..., TW _k=xor (WO _k, TWB _k), wherein, TWB ₁represent the 1st the second bi-level digital watermark keys, TWB ₂represent the 2nd the second bi-level digital watermark keys, TWB _krepresent k the second bi-level digital watermark keys, TWB _krepresent K the second bi-level digital watermark keys, xor () is XOR function;

2.-9, respectively to K bi-level digital watermark TW ₁, TW ₂..., TW _k... and TW _kthe random processing that is inverted, obtains having K bi-level digital watermark of copyright authentication information, and correspondence is designated as respectively

with

2.-10, K bi-level digital watermark of copyright authentication information will be there is

with respectively with multiple zero watermark built-in ends embed K bi-level digital watermark W ₁, W ₂..., W _k... and W _kcorrespondence is carried out similarity calculating one by one, and correspondence obtains K normalized correlation coefficient, then determines whether and extracts K the bi-level digital watermark W that multiple zero watermark built-in ends embed according to the size of K normalized correlation coefficient ₁, W ₂..., W _k... and W _k.

The detailed process of described step 2.-10 is:

Z1, K bi-level digital watermark of copyright authentication information will be there is

with respectively with multiple zero watermark built-in ends embed K bi-level digital watermark W ₁, W ₂..., W _k... and W _kcorrespondence is carried out similarity calculating one by one, and correspondence obtains K normalized correlation coefficient, for inciting somebody to action

with W _kcarry out k the normalized correlation coefficient obtaining after similarity calculating, be designated as

ρ (W_{k}, W_{k}^{*}) = \frac{Σ_{i_{k} = 1}^{I_{k}} Σ_{j_{k} = 1}^{J_{k}} ((w_{k} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}}) \times (w_{k}^{*} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}^{*}}))}{\sqrt{Σ_{i_{k} = 1}^{I_{k}} Σ_{j_{k} = 1}^{J_{k}} {(w_{k} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}})}^{2} Σ_{i_{k} = 1}^{I_{k}} Σ_{j_{k} = 1}^{J_{k}} {(w_{k}^{*} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}^{*}})}^{2}}},

Wherein, w _k(i _k, j _k) represent multiple zero watermark built-in ends embed k bi-level digital watermark W _kmiddle coordinate position is (i _k, j _k) the pixel value of pixel,

k the bi-level digital watermark W that multiple zero watermark built-in ends embed _kin the average of pixel value of all pixels,

represent to have k bi-level digital watermark of copyright authentication information

middle coordinate position is (i _k, j _k) the pixel value of pixel,

in the average of pixel value of all pixels;

Z2, according to the size of K normalized correlation coefficient determine whether extract multiple zero watermark built-in ends embed K bi-level digital watermark W ₁, W ₂..., W _k... and W _k, for normalized correlation coefficient

if

value be 1, determine W _knondestructively extracted, if

value be more than or equal to δ _tand be less than 1, determine W _kextract successfully, if

value be less than δ _t, determine W _kextract unsuccessfully, wherein, δ _trepresent watermark extracting thresholding.

Compared with prior art, the invention has the advantages that:

1) than existing single zero digital watermark, embedding grammar of the present invention can be according to the needs of practical application, when realizing multiple bi-level digital watermark of different objects, different purposes, embed, and do not interfere with each other between the multiple bi-level digital watermarks that embed, therefore more can meet the application diversity of digital watermarking, there is larger actual application value and wide application prospect.

2) than traditional digital watermark technology, embedding grammar of the present invention is not embedded into multiple reality bi-level digital watermarks in original digital image, but be registered in digital watermark information database, therefore not only realized the complete imperceptible of bi-level digital watermark embedding, and raw digital image data is not produced to any destruction, there is not any problem of image quality decrease, maintain original digital image information integrity, this is highly suitable for such as medical image of some responsive digital pictures, the copyright protection of the aspect such as palmprint image and military image.

3) embedding of the present invention and extracting method are to carry out in the compositum of two-dimensional discrete wavelet conversion (DWT) and two-dimension discrete cosine transform (DCT), take full advantage of both features, process multiple zero watermarks of attacking and embed and extract thereby realized the anti-normal image with excellent robust performance, well balance the contradiction between robustness and the imperceptible of digital watermarking.

Accompanying drawing explanation

Fig. 1 a is original Lena digital picture;

Fig. 1 b is the watermark of original two-value signature;

Fig. 1 c is the watermark of original two-value sequence number;

Fig. 1 d is the watermark of original two-value icon;

Fig. 2 a is the Lena digital picture embedding after a binary watermarking;

Fig. 2 b is the Lena digital picture embedding after two binary watermarkings;

Fig. 2 c is the Lena digital picture embedding after three binary watermarkings;

Fig. 2 d is the two-value signature watermark of extracting from Fig. 2 c;

Fig. 2 e is the two-value sequence number watermark of extracting from Fig. 2 c;

Fig. 2 f is the two-value icon watermark of extracting from Fig. 2 c;

Fig. 3 a adds the watermark Lena digital picture after 0.2 to the digital picture shown in Fig. 2 c by all pixel values;

Fig. 3 b adds the watermark Lena digital picture after 0.5 to the digital picture shown in Fig. 2 c by all pixel values;

Fig. 3 c subtracts the watermark Lena digital picture after 0.2 to the digital picture shown in Fig. 2 c by all pixel values;

Fig. 3 d subtracts the watermark Lena digital picture after 0.5 to the digital picture shown in Fig. 2 c by all pixel values;

Fig. 3 e is the two-value signature watermark of extracting from Fig. 3 a, Fig. 3 b and Fig. 3 c;

Fig. 3 f is the two-value sequence number watermark of extracting from Fig. 3 a, Fig. 3 b and Fig. 3 c;

Fig. 3 g is the two-value icon watermark of extracting from Fig. 3 a, Fig. 3 b and Fig. 3 c;

Fig. 4 a is through histogram equalization watermark Lena after treatment digital picture to the digital picture shown in Fig. 2 c;

Fig. 4 b is the two-value signature watermark of extracting from Fig. 4 a;

Fig. 4 c is the two-value sequence number watermark of extracting from Fig. 4 a;

Fig. 4 d is the two-value icon watermark of extracting from Fig. 4 a;

Fig. 5 a is the watermark Lena digital picture after [5 × 5] window medium filtering to the digital picture shown in Fig. 2 c;

Fig. 5 b is the watermark Lena digital picture after [11 × 11] window medium filtering to the digital picture shown in Fig. 2 c;

Fig. 5 c is the two-value signature watermark of extracting from Fig. 5 a;

Fig. 5 d is the two-value sequence number watermark of extracting from Fig. 5 a;

Fig. 5 e is the two-value icon watermark of extracting from Fig. 5 a;

Fig. 5 f is the two-value signature watermark of extracting from Fig. 5 b;

Fig. 5 g is the two-value sequence number watermark of extracting from Fig. 5 b;

Fig. 5 h is the two-value icon watermark of extracting from Fig. 5 b;

Fig. 6 a is the watermark Lena digital picture after the JPEG of 10% quality factor compression to the digital picture shown in Fig. 2 c;

Fig. 6 b is the watermark Lena digital picture after the JPEG of 4% quality factor compression to the digital picture shown in Fig. 2 c;

Fig. 6 c is the two-value signature watermark of extracting from Fig. 6 a;

Fig. 6 d is the two-value sequence number watermark of extracting from Fig. 6 a;

Fig. 6 e is the two-value icon watermark of extracting from Fig. 6 a;

Fig. 6 f is the two-value signature watermark of extracting from Fig. 6 b;

Fig. 6 g is the two-value sequence number watermark of extracting from Fig. 6 b;

Fig. 6 h is the two-value icon watermark of extracting from Fig. 6 b;

Fig. 7 a is to be that 0 variance is the watermark Lena digital picture after 0.02 Gaussian noise to the digital picture stack average shown in Fig. 2 c;

Fig. 7 b is to be that 0 variance is the watermark Lena digital picture after 0.05 Gaussian noise to the digital picture stack average shown in Fig. 2 c;

Fig. 7 c is the two-value signature watermark of extracting from Fig. 7 a;

Fig. 7 d is the two-value sequence number watermark of extracting from Fig. 7 a;

Fig. 7 e is the two-value icon watermark of extracting from Fig. 7 a;

Fig. 7 f is the two-value signature watermark of extracting from Fig. 7 b;

Fig. 7 g is the two-value sequence number watermark of extracting from Fig. 7 b;

Fig. 7 h is the two-value icon watermark of extracting from Fig. 7 b;

Fig. 8 a cuts the watermark Lena digital picture after 128 × 128 pixels of the upper left corner to the digital picture shown in Fig. 2 c;

Fig. 8 b cuts the watermark Lena digital picture after 256 × 256 pixels of the upper left corner to the digital picture shown in Fig. 2 c;

Fig. 8 c is the two-value signature watermark of extracting from Fig. 8 a;

Fig. 8 d is the two-value sequence number watermark of extracting from Fig. 8 a;

Fig. 8 e is the two-value icon watermark of extracting from Fig. 8 a;

Fig. 8 f is the two-value signature watermark of extracting from Fig. 8 b;

Fig. 8 g is the two-value sequence number watermark of extracting from Fig. 8 b;

Fig. 8 h is the two-value icon watermark of extracting from Fig. 8 b;

Fig. 9 a recovers the watermark Lena digital picture after former direction to the digital picture shown in Fig. 2 c again through counterclockwise rotating 5 degree;

Fig. 9 b recovers the watermark Lena digital picture after former direction to the digital picture shown in Fig. 2 c again through counterclockwise rotating 25 degree;

Fig. 9 c is the two-value signature watermark of extracting from Fig. 9 a;

Fig. 9 d is the two-value sequence number watermark of extracting from Fig. 9 a;

Fig. 9 e is the two-value icon watermark of extracting from Fig. 9 a;

Fig. 9 f is the two-value signature watermark of extracting from Fig. 9 b;

Fig. 9 g is the two-value sequence number watermark of extracting from Fig. 9 b;

Fig. 9 h is the two-value icon watermark of extracting from Fig. 9 b.

Embodiment

Below in conjunction with accompanying drawing, embodiment is described in further detail the present invention.

Multiple zero watermark embedding and extracting method of a kind of digital picture that the present invention proposes, watermark embeds and extracts is all to carry out in the compositum of wavelet transform and discrete cosine transform, takes full advantage of both excellent specific properties.The main process that multiple zero watermarks embed is: the wavelet transform that first original digital image is carried out suitable level, and the small echo ll channel obtaining is carried out to discrete cosine transform again, then according to the resolution of K bi-level digital watermark to be embedded, choose respectively the discrete cosine transform ac coefficient that absolute value that K group meets certain difference requires and form K discrete cosine transform ac coefficient sequence, finally construct the bi-level digital watermark keys of K robust according to discrete cosine transform ac coefficient magnitude relationship each other in each sequence, and carry out being saved in registration body after XOR and complete the embedding of K zero watermark with the K being of practical significance a bi-level digital watermark to be embedded, the main process of multiple zero watermark extracting is: by the digital picture of K to be extracted bi-level digital watermark is carried out to similar wavelet transform and discrete cosine transform processing procedure, in conjunction with K zero watermark information being kept in registration body, extract K relevant watermark to prove copyright or the entitlement to original digital image.

Multiple zero watermark embedding methods of a kind of digital picture of the present invention, it comprises the following steps:

1.-1, at multiple zero watermark built-in ends, the gray level image that the original digital image of supposing K to be embedded bi-level digital watermark is 8bit, and be designated as F, F={0≤f (m, n)≤255, 1≤m≤M, 1≤n≤N}, wherein, K >=2, M represents the vertical resolution of the original digital image F of K to be embedded bi-level digital watermark, N represents the lateral resolution of the original digital image F of K to be embedded bi-level digital watermark, M × N represents the resolution of the original digital image F of K to be embedded bi-level digital watermark, f (m, n) represent that in the original digital image F of K to be embedded bi-level digital watermark, coordinate position is (m, the pixel value of pixel n).

1. ,-2, at multiple zero watermark built-in ends, suppose that a K to be embedded digital watermarking is all bianry image, and be designated as W respectively ₁, W ₂..., W _k..., W _k, for k bi-level digital watermark W to be embedded _k, W _k={ w _k(i _k, j _k)=0 or 1,1≤i _k≤ I _k, 1≤j _k≤ J _k, wherein, 1≤k≤K, I _krepresent k bi-level digital watermark W to be embedded _kvertical resolution, J _krepresent k bi-level digital watermark W to be embedded _klateral resolution, I _k× J _krepresent k bi-level digital watermark W to be embedded _kresolution, w _k(i _k, j _k) represent k bi-level digital watermark W to be embedded _kmiddle coordinate position is (i _k, j _k) the pixel value of pixel.

In this specific embodiment, a K to be embedded digital watermarking can be selected has the bianry image such as the corresponding copyright information of practical significance or the author of entitlement information or owner's signature, mark, sequence number, date or company's icon.

1.-3, the original digital image F of K to be embedded bi-level digital watermark is normalized, obtain the digital picture after normalized, be designated as F ', be (m by middle the digital picture F ' after normalized coordinate position, the pixel value of pixel n) is designated as f ' (m, n), f ' (m, n)=f (m, n)/255.

represent to get the maximum integer that is less than himself, this symbol is for rounding symbol, I downwards ₁represent the 1st bi-level digital watermark W to be embedded ₁vertical resolution, J ₁represent the 1st bi-level digital watermark W to be embedded ₁lateral resolution, I _krepresent K bi-level digital watermark W to be embedded _kvertical resolution, J _krepresent K bi-level digital watermark W to be embedded _klateral resolution.

1.-5, FA is carried out to two-dimension discrete cosine transform, obtain one with the first Two Dimension Discrete Cosine matrix of FA equal resolution, be designated as FAC, then FAC being carried out to Zig-Zag scanning arranges, obtain a first one-dimensional discrete cosine transform coefficient sequence, be designated as FACS, FACS={facs (x), 1≤x≤(M/2 ^l) × (N/2 ^l), wherein, facs (x) represents x discrete cosine transform coefficient in FACS, the 2nd discrete cosine transform coefficient in FACS starts to be discrete cosine transform ac coefficient.

1. ,-6, successively according to the resolution of each bi-level digital watermark to be embedded, from FACS, select respectively and meet the I imposing a condition ₁× J ₁, I ₂× J ₂..., I _k× J _k... and I _k× J _kindividual discrete cosine transform ac coefficient, forms K the first one-dimensional discrete cosine transform ac coefficient sequence, and correspondence is designated as FACS respectively ₁, FACS ₂..., FACS _k... and FACS _k, for the satisfied I imposing a condition selecting from FACS _k× J _kthe k that individual discrete cosine transform ac coefficient forms a first one-dimensional discrete cosine transform ac coefficient sequence FACS _k, FACS _k={ facs _k(y), 1≤y≤I _k× J _k, then record FACS ₁, FACS ₂..., FACS _k... and FACS _kin the correspondence position information of various discrete cosine transform ac coefficient in FACS, wherein, I ₂represent the 2nd bi-level digital watermark W to be embedded ₂vertical resolution, J ₂represent the 2nd bi-level digital watermark W to be embedded ₂lateral resolution, facs _k(y) represent FACS _kin y discrete cosine transform ac coefficient, impose a condition and be more than or equal to the difference threshold of setting for the absolute value of difference of two the discrete cosine transform ac coefficients of arbitrary neighborhood in the first one-dimensional discrete cosine transform ac coefficient sequence that makes to form.

At this, choosing of discrete cosine transform ac coefficient in each the first one-dimensional discrete cosine transform ac coefficient sequence is the 2nd discrete cosine transform coefficient (i.e. the 1st discrete cosine transform ac coefficient) since the first one-dimensional discrete cosine transform coefficient sequence FACS, and chooses and the K that forms the first one-dimensional discrete cosine transform ac coefficient sequence FACS ₁, FACS ₂..., FACS _k... and FACS _kin the absolute value of difference of two discrete cosine transform ac coefficients of arbitrary neighborhood be more than or equal to respectively δ ₁, δ ₂..., δ _k... and δ _k, wherein, δ ₁represent for FACS ₁in two discrete cosine transform ac coefficients of arbitrary neighborhood difference absolute value set the 1st difference threshold, δ ₂represent for FACS ₂in two discrete cosine transform ac coefficients of arbitrary neighborhood difference absolute value set the 2nd difference threshold, δ _krepresent for FACS _kin two discrete cosine transform ac coefficients of arbitrary neighborhood difference absolute value set k difference threshold, δ _krepresent for FACS _kin two discrete cosine transform ac coefficients of arbitrary neighborhood difference absolute value set K difference threshold.With δ _kfor example, δ _ksetting principle be such, guaranteeing to choose I _k× J _kunder the prerequisite of individual discrete cosine transform ac coefficient, get maximal value, to improve the unchangeability of digital picture selected mutual magnitude relationship of these discrete cosine transform ac coefficients when under attack the processing.

1.-7, compare respectively FACS ₁, FACS ₂..., FACS _k... and FACS _kin the size of two discrete cosine transform ac coefficients of arbitrary neighborhood, and return to logical value 1 or 0 according to comparative result, for FACS _kin two discrete cosine transform ac coefficients of arbitrary neighborhood, it is designated as respectively to facs _kand facs (z) _k(z+1), judge facs _k(z) >facs _k(z+1) whether set up, if set up, return to logical value 1, otherwise, return to logical value 0, wherein, 1≤z≤I _k× J _k-1; Then compare FACS ₁, FACS ₂..., FACS _k... and FACS _kin last discrete cosine transform ac coefficient and the size of the 1st discrete cosine transform ac coefficient, and return to logical value 1 or 0 according to comparative result, for FACS _kin last discrete cosine transform ac coefficient and the 1st discrete cosine transform ac coefficient, if the former is large, return to logical value 1, otherwise, return to logical value 0; Again according to FACS ₁, FACS ₂..., FACS _k... and FACS _keach self-corresponding logical value of returning, builds the first bi-level digital watermark keys one to one, for FACS _kthe corresponding logical value of returning, it is I that the logical value of returning is stored in to a size by Row Column order arrangement mode _k× J _ktwo-dimensional matrix in, the first bi-level digital watermark keys using this two-dimensional matrix as k, is designated as WB _k.

1.-8, to K to be embedded bi-level digital watermark W ₁, W ₂..., W _k..., W _kcarry out respectively scramble processing, K the bi-level digital watermark obtaining after scramble is processed respectively correspondence is designated as WS ₁, WS ₂..., WS _k... and WS _k, then by WS ₁, WS ₂..., WS _k... and WS _krespectively with K the first bi-level digital watermark keys WB ₁, WB ₂..., WB _k... and WB _kcorrespondence is carried out XOR and is obtained K zero watermark information one by one, and correspondence is designated as WO respectively ₁, WO ₂..., WO _k... and WO _k, WO ₁=xor (WS ₁, WB ₁), WO ₂=xor (WS ₂, WB ₂) ..., WO _k=xor (WS _k, WB _k) ..., WO _k=xor (WS _k, WB _k), then by K zero watermark information WO ₁, WO ₂..., WO _k... and WO _kbe stored in the digital watermark information database of registration body, complete K bi-level digital watermark W ₁, W ₂..., W _k..., W _kembedding, wherein, WS ₁represent W ₁carry out the bi-level digital watermark obtaining after scramble processing, WS ₂represent W ₂carry out the bi-level digital watermark obtaining after scramble processing, WS _krepresent W _kcarry out the bi-level digital watermark obtaining after scramble processing, WS _krepresent W _kcarry out the bi-level digital watermark obtaining after scramble processing, WB ₁represent the 1st the first bi-level digital watermark keys, WB ₂represent the 2nd the first bi-level digital watermark keys, WB _krepresent K the first bi-level digital watermark keys, xor () is XOR function.

At this, to K to be embedded bi-level digital watermark W ₁, W ₂..., W _k..., W _kcarry out respectively scramble processing, can effectively improve the security of zero watermark in registration body, can adopt conventional Arnold conversion in this scramble processing, also can adopt existing other ripe disorder methods.

Multiple zero watermark extracting methods of a kind of digital picture of the present invention, it comprises the following steps:

2.-1, at multiple zero watermark extracting ends, the digital picture of K to be extracted bi-level digital watermark (being test pattern) is designated as to TF, TF={0≤tf (m ', n ')≤255, 1≤m '≤M ', 1≤n '≤N ' }, wherein, K >=2, M ' represents the vertical resolution of the digital picture TF of K to be extracted bi-level digital watermark, N ' represents the lateral resolution of the digital picture TF of K to be extracted bi-level digital watermark, M ' × N ' represents the resolution of the digital picture TF of K to be extracted bi-level digital watermark, the resolution of the digital picture TF of K to be extracted bi-level digital watermark and the embedding of multiple zero watermark built-in ends have the resolution of digital picture of digital watermarking identical, tf (m ', n ') represent in the digital picture TF of K to be extracted bi-level digital watermark coordinate position for (m ', n ') the pixel value of pixel.

2. ,-2, at multiple zero watermark extracting ends, K to be extracted bi-level digital watermark is designated as respectively to W ' ₁, W ' ₂..., W ' _k... and W ' _k, for k bi-level digital watermark W ' to be extracted _k, W ' _k=w ' _k(i ' _k, j ' _k)=0 or 1,1≤i ' _k≤ I ' _k, 1≤j ' _k≤ J ' _k, wherein, 1≤k≤K, I ' _krepresent k bi-level digital watermark W ' to be extracted _kvertical resolution, J ' _krepresent k bi-level digital watermark W ' to be extracted _klateral resolution, I ' _k× J ' _krepresent k bi-level digital watermark W ' to be extracted _kresolution, k bi-level digital watermark W ' to be extracted _kresolution identical with the resolution of k bi-level digital watermark that multiple zero watermark built-in ends embed, w ' _k(i ' _k, j ' _k) represent k bi-level digital watermark W ' to be extracted _kmiddle coordinate position be (i ' _k, j ' _k) the pixel value of pixel.

2.-3, the digital picture TF of K to be extracted bi-level digital watermark is normalized, obtain the digital picture after normalized, be designated as TF ', by middle the digital picture TF ' after normalized coordinate position be (m ', n ') the pixel value of pixel be designated as tf ' (m ', n '), tf ' (m ', n ')=tf (m ', n ')/255.

represent to get the maximum integer that is less than himself, this symbol is for rounding symbol, I downwards ₁' represent the 1st bi-level digital watermark W ' to be extracted ₁vertical resolution, J ₁' represent the 1st bi-level digital watermark W ' to be extracted ₁lateral resolution, I _k' represent K bi-level digital watermark W ' to be extracted _kvertical resolution, J _k' represent K bi-level digital watermark W ' to be extracted _klateral resolution.

2.-5, TFA is carried out to two-dimension discrete cosine transform, obtain one with the second Two Dimension Discrete Cosine matrix of TFA equal resolution, be designated as TFAC, then TFAC being carried out to Zig-Zag scanning arranges, obtain a second one-dimensional discrete cosine transform coefficient sequence, be designated as TFACS, TFACS={tfacs (x '), 1≤x '≤(M '/2 ^l') × (N '/2 ^l'), wherein, tfacs (x ') represents the individual discrete cosine transform coefficient of x ' in TFACS, the 2nd discrete cosine transform coefficient in TFACS starts to be discrete cosine transform ac coefficient.

2.-6, according to the FACS of multiple zero watermark built-in end records ₁, FACS ₂..., FACS _k... and FACS _kin the correspondence position information of various discrete cosine transform ac coefficient in FACS, from TFACS, extract respectively the I of correspondence position ₁' × J ₁', I ₂' × J ₂' ..., I _k' × J _k' ... and I _k' × J _k' individual discrete cosine transform ac coefficient, forms K the second one-dimensional discrete cosine transform ac coefficient sequence, and correspondence is designated as TFACS respectively ₁, TFACS ₂..., TFACS _k... and TFACS _k, for the FACS recording according to multiple zero watermark built-in ends _kin the correspondence position information of various discrete cosine transform ac coefficient in FACS, from TFACS, extract the I of correspondence position _k' × J _kk the second one-dimensional discrete cosine transform ac coefficient sequence TFACS of ' individual discrete cosine transform ac coefficient formation _k, TFACS _k={ tfacs _k(y '), 1≤y '≤I _k' × J _k', wherein, I _k' represent k bi-level digital watermark W ' to be extracted _kvertical resolution, J _k' represent k bi-level digital watermark W ' to be extracted _klateral resolution, tfacs _k(y ') represents TFACS _kin the individual discrete cosine transform ac coefficient of y '.

2.-7, compare respectively TFACS ₁, TFACS ₂..., TFACS _k... and TFACS _kin the size of two discrete cosine transform ac coefficients of arbitrary neighborhood, and return to logical value 1 or 0 according to comparative result, for TFACS _kin two discrete cosine transform ac coefficients of arbitrary neighborhood, it is designated as respectively to tfacs _k(z ') and tfacs _k(z '+1), judges tfacs _k(z ') >tfacs _kwhether set up (z '+1), if set up, returns to logical value 1, otherwise, return to logical value 0, wherein, 1≤z '≤I _k' × J _k'-1; Then compare TFACS ₁, TFACS ₂..., TFACS _k... and TFACS _kin last discrete cosine transform ac coefficient and the size of the 1st discrete cosine transform ac coefficient, and return to logical value 1 or 0 according to comparative result, for TFACS _kin last discrete cosine transform ac coefficient and the 1st discrete cosine transform ac coefficient, if the former is large, return to logical value 1, otherwise, return to logical value 0; Again according to TFACS ₁, TFACS ₂..., TFACS _k... and TFACS _keach self-corresponding logical value of returning, builds the second bi-level digital watermark keys one to one, for TFACS _kthe corresponding logical value of returning, it is I that the logical value of returning is stored in to a size by Row Column order arrangement mode _k' × J _k' two-dimensional matrix in, the second bi-level digital watermark keys using this two-dimensional matrix as k, is designated as TWB _k.

2.-8, by the K from multiple zero watermark built-in ends zero watermark information WO ₁, WO ₂..., WO _k... and WO _k, respectively with K the second bi-level digital watermark keys TWB ₁, TWB ₂..., TWB _k... and TWB _kcorrespondence is carried out XOR one by one, recovers to obtain K bi-level digital watermark, and correspondence is designated as TW respectively ₁, TW ₂..., TW _k... and TW _k, TW ₁=xor (WO ₁, TWB ₁), TW ₂=xor (WO ₂, TWB ₂) ..., TW _k=xor (WO _k, TWB _k) ..., TW _k=xor (WO _k, TWB _k), wherein, TWB ₁represent the 1st the second bi-level digital watermark keys, TWB ₂represent the 2nd the second bi-level digital watermark keys, TWB _krepresent k the second bi-level digital watermark keys, TWB _krepresent K the second bi-level digital watermark keys, xor () is XOR function.

with

with respectively with multiple zero watermark built-in ends embed K bi-level digital watermark W ₁, W ₂..., W _k... and W _kcorrespondence is carried out similarity calculating one by one, and correspondence obtains K normalized correlation coefficient, then determines whether and extracts K the bi-level digital watermark W that multiple zero watermark built-in ends embed according to the size of K normalized correlation coefficient ₁, W ₂..., W _k... and W _kand the robustness of K the bi-level digital watermark embedding.

In this specific embodiment, the detailed process of step 2.-10 is:

with

respectively with multiple zero watermark built-in ends embed K bi-level digital watermark W ₁, W ₂..., W _k... and W _kcorrespondence is carried out similarity calculating one by one, and correspondence obtains K normalized correlation coefficient, for inciting somebody to action

ρ (W_{k}, W_{k}^{*}) = \frac{Σ_{i_{k} = 1}^{I_{k}} Σ_{j_{k} = 1}^{J_{k}} ((w_{k} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}}) \times (w_{k}^{*} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}^{*}}))}{\sqrt{Σ_{i_{k} = 1}^{I_{k}} Σ_{j_{k} = 1}^{J_{k}} {(w_{k} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}})}^{2} Σ_{i_{k} = 1}^{I_{k}} Σ_{j_{k} = 1}^{J_{k}} {(w_{k}^{*} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}^{*}})}^{2}}},

Wherein, w _k(i _k, j _k) represent multiple zero watermark built-in ends embed k bi-level digital watermark W _kmiddle coordinate position is (i _k, j _k) the pixel value of pixel, represent k the bi-level digital watermark W that multiple zero watermark built-in ends embed _kin the average of pixel value of all pixels, represent to have k bi-level digital watermark of copyright authentication information middle coordinate position is (i _k, j _k) the pixel value of pixel,

in the average of pixel value of all pixels.

judgement

value whether be 1, if

value be 1, determine

with W _kin full accord, show W _knondestructively extracted, if this to be Lena digital picture after embed watermark be subject to certain processing or attack after extraction result, the bi-level digital watermark W of embedding be described _kthe ability with this processing of opposing or attack, has desirable robustness.If value be not 1, judge again

value whether between δ _t-1, if value between δ _t-1, the bi-level digital watermark that explanation extracts

bi-level digital watermark W with built-in end embedding _kthere is some difference, but have very large similarity between the two, at this moment can from extract result, recognize the bi-level digital watermark W of embedding _k, extract successfully.The value of related coefficient is larger, the bi-level digital watermark extracting bi-level digital watermark W with built-in end embedding _kmore similar, more easily recognize the bi-level digital watermark W of embedding _k, extraction effect is better.If this to be Lena digital picture after embed watermark be subject to certain processing or attack after extraction result, the bi-level digital watermark W of embedding be described _kthe ability with this processing of more satisfactory opposing or attack, has good robustness.Wherein, δ _tfor watermark extracting thresholding, can value be generally 0.5.If

value not between δ _t-1, judge again value whether be less than δ _t, if so, the bi-level digital watermark that explanation extracts

bi-level digital watermark W with built-in end embedding _kcorrelativity is very little, at this moment cannot from extract result, recognize the bi-level digital watermark W of embedding _k, extract unsuccessfully.

For multiple zero watermark embeddings of a kind of digital picture that the present invention proposes and feasibility and the validity of extracting method are described better, to embed three bi-level digital watermarks as example (K=3), carry out emulation by following experiment.

Experiment simulation carries out on Matlab7.5 platform, and the original digital image F of three bi-level digital watermarks to be embedded selects the Lena gray level image of 8bit, and resolution is 512 × 512, as shown in Figure 1a.Three bi-level digital watermarks are respectively two-value signature watermark W ₁, two-value sequence number watermark W ₂with two-value icon watermark W ₃, resolution is respectively 28 × 50,16 × 64 and 25 × 32, respectively as shown in Fig. 1 b, Fig. 1 c and Fig. 1 d.

Bi-level digital watermark is normalized original digital image F pixel before embedding, and then the digital picture F ' after normalized is carried out to 3 grades of two-dimensional discrete wavelet conversions.

The quality of the digital picture after embed digital watermark adopts Y-PSNR (PSNR) to pass judgment on,

PSNR = - 10 \times \log 10 (\frac{Σ_{m = 1}^{M} Σ_{n = 1}^{N} {(f (m, n) - f^{''} (m, n))}^{2}}{M \times N \times f_{\max}^{2}}),

Wherein, f (m, n) represent that in the original digital image of bi-level digital watermark to be embedded, coordinate position is (m, the pixel value of pixel n), f " (m; n) represent to embed the pixel value of the pixel that in the digital picture after bi-level digital watermark, coordinate position is (m, n), f _maxrepresent the max pixel value of original digital image.

The objective evaluation of digital watermarking testing result adopts normalized correlation coefficient (ρ):

ρ (W_{k}, W_{k}^{*}) = \frac{Σ_{i_{k} = 1}^{I_{k}} Σ_{j_{k} = 1}^{J_{k}} ((w_{k} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}}) \times (w_{k}^{*} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}^{*}}))}{\sqrt{Σ_{i_{k} = 1}^{I_{k}} Σ_{j_{k} = 1}^{J_{k}} {(w_{k} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}})}^{2} Σ_{i_{k} = 1}^{I_{k}} Σ_{j_{k} = 1}^{J_{k}} {(w_{k}^{*} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}^{*}})}^{2}}},

1≤k≤K, wherein, w _k(i _k, j _k) represent k original bi-level digital watermark W _kmiddle coordinate position is (i _k, j _k) the pixel value of pixel, represent k the bi-level digital watermark extracting

middle coordinate position is (i _k, j _k) the pixel value of pixel,

with

representative digit watermark W respectively _kwith

the average of pixel value of all pixels.Can determine whether how extract the bi-level digital watermark of embedding and the bi-level digital watermark robustness of embedding according to the size of normalized correlation coefficient ρ.

Fig. 2 a, Fig. 2 b and Fig. 2 c have provided respectively embedding two-value signature watermark W ₁, embed two-value signature watermark W ₁with two-value sequence number watermark W ₂and embedding two-value signature watermark W ₁, two-value sequence number watermark W ₂with two-value icon watermark W ₃after Lena digital picture.From Fig. 2 a, Fig. 2 b and Fig. 2 c, can see, there is not any variation (PSNR → ∞ dB) in the Lena digital picture quality after embed watermark, consistent with original Lena digital picture, met the requirement of watermark imperceptible completely.Simultaneously visible, utilize multiple zero watermark embedding method of the present invention, the Lena digital picture quality after embed watermark can not become with the watermark number embedding, and therefore can embed easily and flexibly as required in actual applications multiple bi-level digital watermarks.Fig. 2 d, Fig. 2 e and Fig. 2 f are respectively three bi-level digital watermarks that extract from Fig. 2 c, in the time that the watermark Lena digital picture shown in Fig. 2 c is not processed by any attack, can completely nondestructively extract three bi-level digital watermarks of embedding, normalized correlation coefficient ρ is 1.

Watermark Lena digital picture is carried out to various attacks processing below, verify the robustness of multiple zero watermark embedding methods that the present invention proposes.

(1) brightness changes

Watermark Lena digital picture shown in Fig. 2 c is carried out to brightness regulation processing, add respectively 0.2,0.5 and subtract 0.2,0.5 computing by its all pixel values, correspondence obtains the watermark Lena digital picture as shown in Fig. 3 a, Fig. 3 b, Fig. 3 c and Fig. 3 d.After image pixel value plus-minus is processed, vision, there is obvious change in bright, the darkness of the watermark Lena digital picture shown in Fig. 3 a, Fig. 3 b, Fig. 3 c and Fig. 3 d, and Y-PSNR PSNR drops to respectively 13.97dB, 6.02dB, 13.97dB and 6.02dB.Respectively the watermark Lena digital picture of Fig. 3 a, Fig. 3 b, Fig. 3 c and Fig. 3 d is carried out to watermark extracting, three bi-level digital watermarks that extract are respectively as shown in Fig. 3 e, Fig. 3 f and Fig. 3 g.Result demonstration, the watermark result of extracting from these 4 watermark Lena digital pictures is not only identical, and in full accord with three original bi-level digital watermarks.The impact visible, the digital watermarking embedding for the inventive method is not changed by luminance digital image completely, can correctly extract.

(2) histogram equalization

Watermark Lena digital picture shown in Fig. 2 c is carried out to histogram equalization processing, obtain the watermark Lena digital picture as shown in Fig. 4 a.Through histogram equalization processing, the pixel value of watermark Lena digital picture distributes obvious change has occurred, and Y-PSNR PSNR drops to 19.56dB.Fig. 4 b, Fig. 4 c and Fig. 4 d are respectively three bi-level digital watermarks that extract from the watermark Lena digital picture of Fig. 4 a, from result, three bi-level digital watermarks that embed can be extracted very ideally, and normalized correlation coefficient ρ has reached respectively 0.990,0.991 and 0.997.

(3) medium filtering

Watermark Lena digital picture shown in Fig. 2 c is carried out to medium filtering processing, and filtering window size is respectively [5 × 5] and [11 × 11], and the watermark Lena digital picture obtaining after filtering is respectively as shown in Fig. 5 a and Fig. 5 b.Fig. 5 c, Fig. 5 d and Fig. 5 e are respectively three bi-level digital watermarks that extract from the watermark Lena digital picture of Fig. 5 a, and Fig. 5 f, Fig. 5 g and Fig. 5 h are respectively three bi-level digital watermarks that extract from the watermark Lena digital picture of Fig. 5 b.From watermark extracting result, for the medium filtering processing of [5 × 5] wicket, three bi-level digital watermarks that embed are not affected, there is desirable robustness, and for [11 × 11] the medium filtering processing of window greatly, can be seen by Fig. 5 b, at this moment the detailed information of watermark Lena digital picture is very fuzzy, Y-PSNR PSNR drops to 25.83dB, but three bi-level digital watermarks that embed have very desirable anti-filtering processing power, and normalized correlation coefficient ρ has reached respectively 0.964,0.991 and 0.997.

Table 1 has specifically provided watermark Lena digital picture and the watermark extracting result thereof after different windows size medium filtering, can be found out by table, and multiple zero watermark embedding methods that the present invention proposes have ideal anti-filtering processing power.

Watermark Lena digital picture quality after table 1 different windows size medium filtering and extraction result

(4) JPEG lossy compression method

Watermark Lena digital picture shown in Fig. 2 c is carried out to the processing of JPEG lossy compression method, and the compression quality factor is respectively 10% and 4%, and the watermark Lena digital picture obtaining is respectively as shown in Fig. 6 a and Fig. 6 b.Fig. 6 c, Fig. 6 d and Fig. 6 e are respectively three bi-level digital watermarks that extract from the watermark Lena digital picture of Fig. 6 a, and Fig. 6 f, Fig. 6 g and Fig. 6 h are respectively three bi-level digital watermarks that extract from the watermark Lena digital picture of Fig. 6 b.From watermark extracting result, the JPEG lossy compression method processing that is 10% for the compression quality factor, three bi-level digital watermarks that embed are not affected, there is desirable robustness, and the JPEG lossy compression method processing that is 4% for the compression quality factor, can be seen by Fig. 6 b, at this moment watermark Lena digital picture presents very significantly blocking artifact, there is serious degradation in visual quality, Y-PSNR PSNR is only 25.92dB, but three bi-level digital watermarks that embed still have very desirable anti-JPEG lossy compression method processing power, normalized correlation coefficient ρ has reached respectively 0.985, 0.997 and 1.0.

Table 2 has specifically provided watermark Lena digital picture quality and the watermark extracting result thereof under the different JPEG compression quality factors.As can be seen from Table 2, the inventive method has ideal anti-JPEG compression processing power, still can extract three embedded bi-level digital watermarks when the compression quality factor is decreased to 6% zero defect.

Watermark Lena digital picture quality under the different JPEG compression quality of table 2 factor and extraction result

(5) stack Gaussian noise

Watermark Lena digital picture shown in Fig. 2 c is carried out to noise, noise select average be 0 variance be 0.02 and average be that 0 variance is two kinds of Gaussian noises of 0.05, the watermark Lena digital picture obtaining is respectively as shown in Fig. 7 a and Fig. 7 b.Fig. 7 c, Fig. 7 d and Fig. 7 e are respectively three bi-level digital watermarks that extract from the watermark Lena digital picture of Fig. 7 a, and Fig. 7 f, Fig. 7 g and Fig. 7 h are respectively three bi-level digital watermarks that extract from the watermark Lena digital picture of Fig. 7 b.From watermark extracting result, although watermark Lena digital picture is subject to Gauusian noise jammer, visual quality generation serious degradation, Y-PSNR PSNR is only 17.22dB and 13.67dB.But be that 0 variance is 0.02 Gauusian noise jammer for average, three bi-level digital watermarks that embed are not affected, there is desirable robustness, and be that 0 variance is 0.05 Gauusian noise jammer for average, three bi-level digital watermarks that embed still have very desirable anti-noise jamming ability, and normalized correlation coefficient ρ has reached respectively 0.980,0.994 and 1.0.

It is watermark Lena digital picture quality and the watermark extracting result thereof under 0 different variance Gauusian noise jammers that table 3 has specifically provided average.As can be seen from Table 3, the inventive method has ideal anti-noise jamming ability, the Gauusian noise jammer that is less than 0.03 for variance, and three bi-level digital watermarks of embedding can be extracted intactly.

Watermark Lena digital picture quality under the different Gaussian noise intensity of table 3 and extraction result

(6) how much cuttings

Watermark Lena digital picture shown in Fig. 2 c is carried out to geometry cutting, start from the upper left corner to cut respectively 128 × 128 and 256 × 256 pixels, obtain the watermark Lena digital picture as shown in Fig. 8 a and Fig. 8 b.Fig. 8 c, Fig. 8 d and Fig. 8 e are respectively three bi-level digital watermarks that extract from the watermark Lena digital picture of Fig. 8 a, and Fig. 8 f, Fig. 8 g and Fig. 8 h are respectively three bi-level digital watermarks that extract from the watermark Lena digital picture of Fig. 8 b.From watermark extracting result, although watermark Lena digital picture is destroyed largely, Y-PSNR PSNR is only 17.29dB and 11.14dB, but the inventive method has goodish robustness for how much cuttings, three bi-level digital watermarks that embed still can be by intact extracting very, and normalized correlation coefficient ρ has reached respectively 0.973,0.972 and 0.981 and 0.874,0.883 and 0.899.

(7) how much rotations

Watermark Lena digital picture shown in Fig. 2 c is carried out to counter clockwise direction rotation, and angle is respectively 5 degree and 25 degree.In order to extract watermark, by the reverse rotation again of postrotational image, to recover former direction, the watermark Lena digital picture at this moment obtaining is respectively as shown in Fig. 9 a and Fig. 9 b, and its Y-PSNR PSNR is 19.10dB and 13.80dB.Fig. 9 c, Fig. 9 d and Fig. 9 e are respectively three bi-level digital watermarks that extract from the watermark Lena digital picture of Fig. 9 a, and Fig. 9 f, Fig. 9 g and Fig. 9 h are respectively three bi-level digital watermarks that extract from the watermark Lena digital picture of Fig. 9 b.From watermark extracting result, the inventive method also has reasonable robustness for how much rotation attacks, three bi-level digital watermarks that embed still can very clearly be extracted, and normalized correlation coefficient ρ has reached respectively 0.819,0.829 and 0.871 and 0.694,0.751 and 0.817.

Claims

1. multiple zero watermark embedding methods for digital picture, is characterized in that comprising the following steps:

1. ,-2, at multiple zero watermark built-in ends, suppose that K to be embedded bi-level digital watermark is all bianry image, and be designated as W respectively ₁, W ₂..., W _k..., W _k, for k bi-level digital watermark W to be embedded _k, W _k={ w _k(i _k, j _k)=0 or 1,1≤i _k≤ I _k, 1≤j _k≤ J _k, wherein, 1≤k≤K, I _krepresent k bi-level digital watermark W to be embedded _kvertical resolution, J _krepresent k bi-level digital watermark W to be embedded _klateral resolution, I _k× J _krepresent k bi-level digital watermark W to be embedded _kresolution, w _k(i _k, j _k) represent k bi-level digital watermark W to be embedded _kmiddle coordinate position is (i _k, j _k) the pixel value of pixel;

1.-3, the original digital image F of K to be embedded bi-level digital watermark is normalized, obtain the digital picture after normalized, be designated as F', be (m by coordinate position in the digital picture F' after normalized, the pixel value of pixel n) is designated as f'(m, n), f'(m, n)=f (m, n)/255;

1.-4, F' is carried out to L level two-dimensional discrete wavelet conversion, obtain a first small echo ll channel and multiple the first wavelet details subgraph, the first small echo ll channel is designated as to FA, wherein, the resolution of FA is (M/2 ^l) × (N/2 ^l),

2. multiple zero watermark embedding methods of a kind of digital picture according to claim 1, is characterized in that FACS in described step 1.-6 ₁, FACS ₂..., FACS _k... and FACS _kin the absolute value of difference of two discrete cosine transform ac coefficients of arbitrary neighborhood be more than or equal to respectively δ ₁, δ ₂..., δ _k... and δ _k, wherein, δ ₁represent for FACS ₁in two discrete cosine transform ac coefficients of arbitrary neighborhood difference absolute value set the 1st difference threshold, δ ₂represent for FACS ₂in two discrete cosine transform ac coefficients of arbitrary neighborhood difference absolute value set the 2nd difference threshold, δ _krepresent for FACS _kin two discrete cosine transform ac coefficients of arbitrary neighborhood difference absolute value set k difference threshold, δ _krepresent for FACS _kin two discrete cosine transform ac coefficients of arbitrary neighborhood difference absolute value set K difference threshold.

3. multiple zero watermark extracting methods for the digital picture corresponding with multiple zero watermark embedding methods of digital picture claimed in claim 1, is characterized in that comprising the following steps:

2.-1, at multiple zero watermark extracting ends, the digital picture of K to be extracted bi-level digital watermark is designated as to TF, TF={0≤tf (m', n')≤255, 1≤m'≤M', 1≤n'≤N'}, wherein, K >=2, M' represents the vertical resolution of the digital picture TF of K to be extracted bi-level digital watermark, N' represents the lateral resolution of the digital picture TF of K to be extracted bi-level digital watermark, M' × N' represents the resolution of the digital picture TF of K to be extracted bi-level digital watermark, the resolution of the digital picture TF of K to be extracted bi-level digital watermark and the embedding of multiple zero watermark built-in ends have the resolution of digital picture of digital watermarking identical, tf (m', n') represent that in the digital picture TF of K to be extracted bi-level digital watermark, coordinate position is (m', the pixel value of pixel n'),

2. ,-2, at multiple zero watermark extracting ends, K to be extracted bi-level digital watermark is designated as respectively to W ₁', W ₂' ..., W _k' ... and W' _k, for k bi-level digital watermark W' to be extracted _k, W' _k={ w' _k(i' _k, j' _k)=0 or 1,1≤i' _k≤ I' _k, 1≤j' _k≤ J' _k, wherein, 1≤k≤K, I' _krepresent k bi-level digital watermark W' to be extracted _kvertical resolution, J' _krepresent k bi-level digital watermark W to be extracted _k' lateral resolution, I' _k× J' _krepresent k bi-level digital watermark W to be extracted _k' resolution, k bi-level digital watermark W to be extracted _k' resolution identical with the resolution of k bi-level digital watermark that multiple zero watermark built-in ends embed, w' _k(i' _k, j' _k) represent k bi-level digital watermark W' to be extracted _kmiddle coordinate position is (i' _k, j' _k) the pixel value of pixel;

2.-3, the digital picture TF of K to be extracted bi-level digital watermark is normalized, obtain the digital picture after normalized, be designated as TF', be (m' by coordinate position in the digital picture TF' after normalized, the pixel value of pixel n') is designated as tf'(m', n'), tf'(m', n')=tf (m', n')/255;

2.-4, TF' is carried out to L' level two-dimensional discrete wavelet conversion, obtain a second small echo ll channel and multiple the second wavelet details subgraph, the second small echo ll channel is designated as to TFA, wherein, the resolution of TFA is (M'/2 ^l') × (N'/2 ^l'),

represent to get the maximum integer that is less than himself, I ₁' represent the 1st bi-level digital watermark W to be extracted ₁' vertical resolution, J ₁' represent the 1st bi-level digital watermark W to be extracted ₁' lateral resolution, I _k' represent K bi-level digital watermark W to be extracted _k' vertical resolution, J _k' represent K bi-level digital watermark W to be extracted _k' lateral resolution;

2.-5, TFA is carried out to two-dimension discrete cosine transform, obtain one with the second Two Dimension Discrete Cosine matrix of TFA equal resolution, be designated as TFAC, then TFAC being carried out to Zig-Zag scanning arranges, obtain a second one-dimensional discrete cosine transform coefficient sequence, be designated as TFACS, TFACS={tfacs (x'), 1≤x'≤(M'/2 ^l') × (N'/2 ^l'), wherein, tfacs (x') represents x' discrete cosine transform coefficient in TFACS, the 2nd discrete cosine transform coefficient in TFACS starts to be discrete cosine transform ac coefficient;

2.-6, according to the FACS of multiple zero watermark built-in end records ₁, FACS ₂..., FACS _k... and FACS _kin the correspondence position information of various discrete cosine transform ac coefficient in FACS, from TFACS, extract respectively the I of correspondence position ₁' × J ₁', I ₂' × J ₂' ..., I _k' × J _k' ... and I _k' × J _k' individual discrete cosine transform ac coefficient, forming K the second one-dimensional discrete cosine transform ac coefficient sequence, correspondence is designated as TFACS respectively ₁, TFACS ₂..., TFACS _k... and TFACS _k, for the FACS recording according to multiple zero watermark built-in ends _kin the correspondence position information of various discrete cosine transform ac coefficient in FACS, from TFACS, extract the I of correspondence position _k' × J _k' k the second one-dimensional discrete cosine transform ac coefficient sequence TFACS forming of individual discrete cosine transform ac coefficient _k, TFACS _k={ tfacs _k(y'), 1≤y'≤I _k' × J _k', wherein, I _k' represent k bi-level digital watermark W to be extracted _k' vertical resolution, J _k' represent k bi-level digital watermark W to be extracted _k' lateral resolution, tfacs _k(y') represent TFACS _kin y' discrete cosine transform ac coefficient;

2.-7, compare respectively TFACS ₁, TFACS ₂..., TFACS _k... and TFACS _kin the size of two discrete cosine transform ac coefficients of arbitrary neighborhood, and return to logical value 1 or 0 according to comparative result, for TFACS _kin two discrete cosine transform ac coefficients of arbitrary neighborhood, it is designated as respectively to tfacs _kand tfacs (z') _k(z'+1), judge tfacs _k(z') >tfacs _k(z'+1) whether set up, if set up, return to logical value 1, otherwise, return to logical value 0, wherein, 1≤z'≤I _k' × J _k'-1; Then compare TFACS ₁, TFACS ₂..., TFACS _k... and TFACS _kin last discrete cosine transform ac coefficient and the size of the 1st discrete cosine transform ac coefficient, and return to logical value 1 or 0 according to comparative result, for TFACS _kin last discrete cosine transform ac coefficient and the 1st discrete cosine transform ac coefficient, if the former is large, return to logical value 1, otherwise, return to logical value 0; Again according to TFACS ₁, TFACS ₂..., TFACS _k... and TFACS _keach self-corresponding logical value of returning, builds the second bi-level digital watermark keys one to one, for TFACS _kthe corresponding logical value of returning, it is I that the logical value of returning is stored in to a size by Row Column order arrangement mode _k' × J _k' two-dimensional matrix in, the second bi-level digital watermark keys using this two-dimensional matrix as k, is designated as TWB _k;

2.-9, respectively to K bi-level digital watermark TW ₁, TW ₂..., TW _k... and TW _kthe random processing that is inverted, obtains having K bi-level digital watermark of copyright authentication information, and correspondence is designated as W respectively ₁ ^*, W ₂ ^*..., W _k ^*... with

2.-10, K bi-level digital watermark W of copyright authentication information will be there is ₁ ^*, W ₂ ^*..., W _k ^*... with

respectively with multiple zero watermark built-in ends embed K bi-level digital watermark W ₁, W ₂..., W _k... and W _kcorrespondence is carried out similarity calculating one by one, and correspondence obtains K normalized correlation coefficient, then determines whether and extracts K the bi-level digital watermark W that multiple zero watermark built-in ends embed according to the size of K normalized correlation coefficient ₁, W ₂..., W _k... and W _k.

4. multiple zero watermark extracting methods of a kind of digital picture according to claim 3, is characterized in that the detailed process of described step 2.-10 is:

Z1, K bi-level digital watermark W of copyright authentication information will be there is ₁ ^*, W ₂ ^*..., W _k ^*... with

respectively with multiple zero watermark built-in ends embed K bi-level digital watermark W ₁, W ₂..., W _k... and W _kcorrespondence is carried out similarity calculating one by one, and correspondence obtains K normalized correlation coefficient, for by W _k ^*with W _kcarry out k the normalized correlation coefficient obtaining after similarity calculating, be designated as ρ (W _k, W _k ^*),

ρ (W_{k}, {W_{k}}^{*}) = \frac{Σ_{i_{k} = 1}^{I_{k}} Σ_{j_{k} = 1}^{J_{k}} ((w_{k} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}}) \times (w_{k}^{*} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}^{*}}))}{\sqrt{Σ_{i_{k} = 1}^{I_{k}} Σ_{j_{k} = 1}^{J_{k}} {(w_{k} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}})}^{2} \times Σ_{i_{k} = 1}^{I_{k}} Σ_{j_{k} = 1}^{J_{k}} {(w_{k}^{*} (i_{k}, j_{k}) - \overset{&OverBar;}{w_{k}^{*}})}^{2}}},

represent k the bi-level digital watermark W that multiple zero watermark built-in ends embed _kin the average of pixel value of all pixels,

represent to have k bi-level digital watermark W of copyright authentication information _k ^*middle coordinate position is (i _k, j _k) the pixel value of pixel,

represent to have k bi-level digital watermark W of copyright authentication information _k ^*in the average of pixel value of all pixels;

Z2, according to the size of K normalized correlation coefficient determine whether extract multiple zero watermark built-in ends embed K bi-level digital watermark W ₁, W ₂..., W _k... and W _k, for normalized correlation coefficient ρ (W _k, W _k ^*), if ρ is (W _k, W _k ^*) value be 1, determine W _knondestructively extracted, if ρ is (W _k, W _k ^*) value be more than or equal to δ _tand be less than 1, determine W _kextract successfully, if ρ is (W _k, W _k ^*) value be less than δ _t, determine W _kextract unsuccessfully, wherein, δ _trepresent watermark extracting thresholding.