CN102883091A - Digital watermark embedding method and extracting method for image retrieval and copyright protection - Google Patents

Abstract

The invention discloses a digital watermark embedding method and a digital watermark extracting method for image retrieval and copyright protection. The digital watermark embedding method introduces a digital watermark technology to the process of image retrieval, and a binary digital watermark is embedded into a digital image in a digital image work warehousing link, so that a digital image work is under the copyright protection, and the safety of the digital image work in the process of digital image retrieval is improved; the digital watermark extracting method introduces the digital watermark technology to the process of image retrieval, the embedded binary digital watermark is used as query information during the digital image retrieval, and the extracted binary digital watermark is used as feature information during the digital image retrieval, so that retrieval pertinence is ensured, the establishment of a high-capacity index database is avoided, and a storage space is saved; and the embedded watermark has good robustness, therefore after the digital image work is subjected to the conventional image processing or is under external attack, the embedded watermark can still be effectively extracted, and the accuracy of the image retrieval is ensured.

Description

A kind of digital watermarking embedding and extracting method for image retrieval and copyright protection

Technical field

The present invention relates to a kind of multimedia information retrieval and information security technology, especially relate to a kind of digital watermarking embedding and extracting method for image retrieval and copyright protection.

Background technology

Along with developing rapidly of multimedia technology, computer technology and network technology, various types of digital multimedia information produce continuously, transmit and are applied.Digital picture as a kind of abundant in content, express intuitively multimedia messages, its be the most important in the multimedia information lib also be to use the most general a kind of Digital Media.At present, the data of digital picture just with the speed of explosion type in sharp increase.Obviously, how the mass digital view data that produces facing to sharp increase is carried out effective organization and management, and is therefrom retrieved rapidly the image information that meets user's request, has become the current realistic problem that will solve of needing badly.

From the mass digital view data, retrieve rapidly the image information that meets user's request, developed text based and two kinds of content-based digital image search technology at present.Text based digital image search technology is namely to set up corresponding keyword or description field by digital picture is carried out text marking, critical information such as the time of title, numbering, author, content description, image size, the image of digital picture being originated, setting up image, storage location adopts automatic indexing or carries out manual annotation, set up the keyword index database of digital picture, then press the full-text data library management, adopt the signature method to realize.Therefore, be a kind of matched and searched process based on text key word on the text based digital image search technical spirit, namely the digital image search process is converted to the text retrieval corresponding with this digital picture.Content-based digital image search technology refers to search the digital picture that contains specific objective from image library, it carries out automatic analysis by software or corresponding tool to digital picture, extract the content characteristic of digital picture, such as color, shape and texture etc., and the combination of these features, deposit the aspect indexing database of digital picture in as characteristic vector, like this when carrying out digital image search, a given retrieving images is carried out graphical analysis extract characteristic vector, then utilize similarity matching methods to calculate the characteristic vector of retrieving images and the similarity of each image feature vector in the aspect indexing database, return result for retrieval according to the size of similarity.

Yet, being accompanied by further developing of digital image search technology, new problem is also following.At first, based on text and two kinds of content-based digital image search technology, in retrieving, all need to set up jumbo index data base.Along with the sharp increase of DID amount, be a very large challenge for memory space; Secondly, because constantly improving and development of network technology, interconnected between system becomes very easy, although open internet is very easy to people's obtaining, exchanging and sharing for digital image information, but correspondingly, the safety issue of the aspects such as the copyright of digital picture and content integrity also becomes and becomes increasingly conspicuous.For example, any user can retrieve the digital picture works that obtain other people rapidly by the internet, if in the situation that do not obtain author or copyright owner's license, original digital image is arbitrarily copied, revised and again transmission, will be arrived author's copyright and copyright owner's interests by serious infringement like this.Study hotspot as Contemporary Digital media information security fields; digital watermark technology has been widely used in copyright protection and the content integrity authentication of digital picture; it carries out certain modification by spatial domain pixel or the coefficient in transform domain to original digital image; to comprise certain or several characteristic information that author or copyright owner have the authentication of copyright protection and content integrity for example signs; copyright marking; sequence number; date or icon etc. are embedded in the original digital image as watermark information, be reached for the original digital image works with this authentication information of necessity and the purpose of copyright protection are provided.Digital watermarking will be played due effect, and General Requirements possesses robustness and two fundamentals of imperceptible.Robustness refer to original digital image process through normal signal or external attack after, the digital watermarking of embedding still has the information that detectability preferably still can reflect the aspects such as copyright of original digital image in other words.Imperceptible refers to that embedding algorithm can not have influence on the visual quality of original digital image, thereby can not affect the using value of original digital image.

Therefore; in the development result that takes full advantage of multimedia technology, computer technology and network technology; enjoy the DID resource of magnanimity and when therefrom quick-searching goes out to meet the image information of user's request; if can effectively protect the integrality of these raw digital image datas and the aspect problems such as copyright of works again, it will be the common development direction of multimedia information retrieval and field of information security technology.

Summary of the invention

It is good that technical problem to be solved by this invention provides a kind of watermark robustness, the data waterprint embedded method of the fail safe of works in can Effective Raise digital image search process, and a kind of accuracy that can the Effective Raise digital image search, and can effectively solve the digital watermarking extracting method of memory space problem.

The present invention solves the problems of the technologies described above the technical scheme that adopts: a kind of data waterprint embedded method for image retrieval and copyright protection, the processing procedure that it is characterized in that it is: at first, the original figure 8bit gray level image of bi-level digital watermark to be embedded is carried out normalized; Secondly, the digital picture that obtains after the normalized is carried out two-dimensional discrete wavelet conversion, obtain first a small echo ll channel and a plurality of the first wavelet details subgraph; The first small echo ll channel is decomposed into a fractional part the first small echo ll channel and integer part first a small echo ll channel; Again, pixel value according to each pixel in the bi-level digital watermark to be embedded, coefficient value scope in conjunction with each coefficient in fractional part the first small echo ll channel, coefficient value to each coefficient in fractional part the first small echo ll channel carries out the two-value quantification treatment, realize the embedding of bi-level digital watermark, and return a two-valued function table; Then, new fractional part the first small echo ll channel and integer part the first small echo ll channel that obtains after the two-value quantification treatment is combined into the first new small echo ll channel; The first new small echo ll channel and a plurality of the first wavelet details subgraph are carried out the 2-d discrete wavelet inverse transformation together, obtain embedding the digital picture of bi-level digital watermark; At last, the bi-level digital watermark is reached the logical table that in the bi-level digital watermark embed process, returns and be transferred to digital watermarking extraction end.

A kind of data waterprint embedded method for image retrieval and copyright protection of the present invention, it specifically may further comprise the steps:

1.-1, at the digital watermarking built-in end, suppose to have K bi-level digital watermark to be embedded and the original figure 8bit gray level image of K bi-level digital watermark to be embedded, k bi-level digital watermark is designated as W _k, W _k={ w _k| w _k(m _k, n _k)=0 or 1,1≤m _k≤ M _k, 1≤n _k≤ N _k, the original figure 8bit gray level image of k bi-level digital watermark to be embedded is designated as F _k, F _k={ f _k| 0≤f _k(i _k, j _k)≤255,1≤i _k≤ I _k, 1≤j _k≤ J _k, wherein, 1≤k≤K, K 〉=1, W _kResolution be M _k* N _k, at this hypothesis M _k=N _k, w _k(m _k, n _k) expression W _kMiddle coordinate position is (m _k, n _k) the pixel value of pixel, F _kResolution be I _k* J _k, at this hypothesis I _k=J _k, f _k(i _k, j _k) expression F _kMiddle coordinate position is (i _k, j _k) the pixel value of pixel;

1.-2 the original figure 8bit gray level image of, supposing the current bi-level digital watermark to be embedded of processing is the original figure 8bit gray level image F of k bi-level digital watermark to be embedded _k

1.-3, to F _kCarry out normalized, with F _kThe digital picture F ' that after normalized, obtains _kMiddle coordinate position is (i _k, j _k) the pixel value of pixel be designated as f ' _k(i _k, j _k), f ' _k(i _k, j _k)=f _k(i _k, j _k)/255, wherein, 1≤i _k≤ I _k, 1≤j _k≤ J _k

1.-4, to F ' _kCarry out L _kThe level two-dimensional discrete wavelet conversion obtains first a small echo ll channel and a plurality of the first wavelet details subgraph, with F ' _kThe first corresponding small echo ll channel is designated as FA _k, wherein, L _k=log ₂(I _k/ M _k)=log ₂(J _k/ N _K), FA _kResolution be

1.-5, with FA _kBe decomposed into fractional part and integer part two parts data subgraph, obtain a fractional part the first small echo ll channel and integer part first a small echo ll channel, be designated as respectively FAD _kAnd FAI _k

1.-6, according to W _kIn the pixel value of each pixel, to FAD _kIn the coefficient value of each coefficient carry out the two-value quantification treatment, realize bi-level digital watermark W _kEmbedding, and to return a size be M _k* N _kThe two-valued function table, the fractional part first small echo ll channel that will obtain after the two-value quantification treatment is designated as FAD ' _k, be Lg with the two-valued function souvenir that returns _k

1.-7, with FAD ' _kAnd FAI _kBe combined into F ' _kCorresponding the first new small echo ll channel is designated as FA ' _k

1.-8, with FA ' _kAnd F ' _kCorresponding a plurality of the first wavelet details subgraphs carry out L together again _kThe inverse transformation of level 2-d discrete wavelet, obtaining embedding has bi-level digital watermark W _kDigital picture, be designated as F " _k

1.-9, make k=k+1, then returning step 1.-2 continues the original figure 8bit gray level image of next bi-level digital watermark to be embedded is processed, until the original figure 8bit gray level image of K bi-level digital watermark to be embedded all is disposed, wherein, "=" among the k=k+1 is assignment;

1.-10, with K bi-level digital watermark W ₁, W ₂..., W _k..., W _KAnd the two-valued function table Lg that in the bi-level digital watermark embed process, returns ₁, Lg ₂..., Lg _k..., Lg _KBe transferred to digital watermarking and extract end, wherein, W ₁Represent the 1st bi-level digital watermark, W ₂Represent the 2nd bi-level digital watermark, W _KRepresent K bi-level digital watermark, Lg ₁Be illustrated in bi-level digital watermark W ₁The two-valued function table that returns in the telescopiny, Lg ₂Be illustrated in bi-level digital watermark W ₂The two-valued function table that returns in the telescopiny, Lg _KBe illustrated in bi-level digital watermark W _KThe two-valued function table that returns in the telescopiny.

Described step 1.-5 in the order

Expression FA _kMiddle coordinate position is

The coefficient value of coefficient, order Expression FAD _kMiddle coordinate position is

The coefficient value of coefficient, order

Expression FAI _kMiddle coordinate position is

The coefficient value of coefficient, then have

Wherein, symbol

The maximum integer less than himself is got in expression, $1\≤\widetilde{i_k}\≤(I_k/2^{L_k}),$ $1\≤\widetilde{j_k}\≤(J_k/2^{L_k}).$

Described step detailed process 1.-6 is: a1, with FAD _kIn the current Coefficient Definition of processing be current coefficient; A2, suppose that the coordinate position of current coefficient is (m _k, n _k), then the coefficient value with current coefficient is designated as fad _k(m _k, n _k), wherein, 1≤m _k≤ M _k, 1≤n _k≤ N _kA3, according to fad _k(m _k, n _k) and W _kMiddle coordinate position is (m _k, n _k) the pixel value w of pixel _k(m _k, n _k), calculate FAD ' _kMiddle coordinate position is (m _k, n _k) the coefficient value of coefficient, be designated as fad ' _k(m _k, n _k), and return two-valued function table Lg _kMiddle coordinate position is (m _k, n _k) the logical value lg of element _k(m _k, n _k), specific as follows: as to work as w _k(m _k, n _k)=0 and fad _k(m _k, n _k) ∈ [0,0.5) time, with fad _k(m _k, n _k) be quantified as 0.25, i.e. fad ' _k(m _k, n _k)=0.25, lg _k(m _k, n _k)=0; Work as w _k(m _k, n _k)=0 and fad _k(m _k, n _k) ∈ [0.5,1), with fad _k(m _k, n _k) be quantified as 0.75, i.e. fad ' _k(m _k, n _k)=0.75, lg _k(m _k, n _k)=1; Work as w _k(m _k, n _k)=1 and fad _k(m _k, n _k) ∈ [0,0.5), with fad _k(m _k, n _k) be quantified as 0.25, i.e. fad ' _k(m _k, n _k)=0.25, lg _k(m _k, n _kW is worked as in)=1 _k(m _k, n _k)=1 and fad _k(m _k, n _k) ∈ [0.5,1), with fad _k(m _k, n _k) be quantified as 0.75, i.e. fad ' _k(m _k, n _k)=0.75, lg _k(m _k, n _k)=0; A4, with FAD _kThen the pending coefficient of the middle next one returns step a2 and continues to carry out, until FAD as current coefficient _kIn all coefficient processing complete, realize bi-level digital watermark W _kEmbedding, obtain fractional part the first small echo ll channel FAD ' that after the two-value quantification treatment, obtains _k

Described step detailed process 1.-7 is: b1, with FAD ' _kIn the current Coefficient Definition of processing be current the first coefficient, with FAI _kIn the current Coefficient Definition of processing be current the second coefficient; B2, suppose that the coordinate position of current the first coefficient is

Then the coefficient value with current the first coefficient is designated as

The coordinate position of supposing current the second coefficient is

Then the coefficient value with current the second coefficient is designated as

Wherein,

B3, basis

With

Calculate FA ' _kMiddle coordinate position is The coefficient value of coefficient, be designated as

${\mathrm{fa}}_k^{\′}(\widetilde{i_k},\widetilde{j_k})={\mathrm{fad}}_k^{\′}(\widetilde{i_k},\widetilde{j_k})+f{\mathrm{ai}}_k(\widetilde{i_k},\widetilde{j_k});$ B4, with FAD ' _kThe pending coefficient of the middle next one is as current the first coefficient, with FAI _kThen the pending coefficient of the middle next one returns step b2 and continues to carry out, until FAD ' as current the second coefficient _kAnd FAI _kIn all coefficient processing complete, obtain F ' _kCorresponding the first new small echo ll channel FA ' _k

A kind of digital watermarking extracting method for image retrieval and copyright protection is characterized in that its processing procedure is: at first, the digital 8bit gray level image of bi-level digital watermark to be extracted is carried out normalized; Secondly, the digital picture that obtains after the normalized is carried out two-dimensional discrete wavelet conversion, obtain second a small echo ll channel and a plurality of the second wavelet details subgraph; The second small echo ll channel is decomposed into a fractional part the second small echo ll channel and integer part second a small echo ll channel; Again, be transferred to the two-valued function table that end is extracted in digital watermarking according to the digital watermarking built-in end, and the coefficient value scope of each coefficient in fractional part the second small echo ll channel, carry out two-value and differentiate, extract the bi-level digital watermark; Then, be transferred to the coefficient correlation that bi-level digital watermark with the bi-level digital watermark that extracts of end are extracted in digital watermarking by calculating the digital watermarking built-in end, determine whether the bi-level digital watermark that extracts is the bi-level digital watermark that is used for copyright protection that the digital watermarking built-in end embeds; At last, utilize coefficient correlation to realize the retrieval of digital picture.

A kind of digital watermarking extracting method for image retrieval and copyright protection of the present invention, it specifically may further comprise the steps:

2.-1, extract end in digital watermarking, suppose to have the individual bi-level digital watermark to be extracted of K and the digital 8bit gray level image of K bi-level digital watermark to be extracted, k bi-level digital watermark is designated as W ' _k, W ' _k=w ' _k| w ' _k(m ' _k, n ' _k)=0 or 1,1≤m ' _k≤ M ' _k, 1≤n ' _k≤ N ' _k, the digital 8bit gray level image of k bi-level digital watermark to be extracted is designated as TF _k, TF _k={ tf _k| 0≤tf _k(i ' _k, j ' _k)≤255,1≤i ' _k≤ I ' _k, 1≤j ' _k≤ J ' _k, wherein, 1≤k≤K, K 〉=1, W ' _kResolution be M ' _k* N ' _k, at this hypothesis M ' _k=N ' _k, and W ' _kThe resolution of the bi-level digital watermark that embeds of resolution and digital watermarking built-in end identical, w ' _k(m ' _k, n ' _k) expression W ' _kMiddle coordinate position be (m ' _k, n ' _k) the pixel value of pixel, TF _kResolution be I ' _k* J ' _k, at this hypothesis I ' _k=J ' _k, and TF _kResolution identical with the resolution of the original figure 8bit gray level image of digital watermarking built-in end bi-level digital watermark to be embedded, tf _k(i ' _k, j ' _k) expression TF _kMiddle coordinate position be (i ' _k, j ' _k) the pixel value of pixel;

2.-2 the digital 8bit gray level image of, supposing the current bi-level digital watermark to be extracted of processing is the digital 8bit gray level image TF of k bi-level digital watermark to be extracted _k

2.-3, to TF _kCarry out normalized, with TF _kThe digital picture TF ' that after normalized, obtains _kMiddle coordinate position be (i ' _k, j ' _k) the pixel value of pixel be designated as tf ' _k(i ' _k, j ' _k), tf ' _k(i ' _k, j ' _k)=tf _k(i ' _k, j ' _k)/255, wherein, 1≤i ' _k≤ I ' _k, 1≤j ' _k≤ J ' _k

2.-4, to TF ' _kCarry out L ' _kThe level two-dimensional discrete wavelet conversion obtains second a small echo ll channel and a plurality of the second wavelet details subgraph, with TF ' _kThe second corresponding small echo ll channel is designated as TFA _k, wherein, L ' _k=log ₂(I ' _k/ M ' _k)=log ₂(J ' _k/ N ' _k), TFA _kResolution be

2.-5, with TFA _kBe decomposed into fractional part and integer part two parts data subgraph, obtain a fractional part the second small echo ll channel and integer part second a small echo ll channel, be designated as respectively TFAD _kAnd TFAI _k

2.-6, be transferred to the two-valued function table Lg that end is extracted in digital watermarking according to the digital watermarking built-in end _kAnd TFAD _kIn the coefficient value of each coefficient, extract bi-level digital watermark W ' _k

2.-7, calculate the digital watermarking built-in end and be transferred to the bi-level digital watermark W that end is extracted in digital watermarking _kWith the bi-level digital watermark W ' that extracts _kNormalizated correlation coefficient, be designated as r, r=corr (W _k, W ' _k), then determine the bi-level digital watermark W that is used for copyright protection that the digital watermarking built-in end embeds according to the size of r _kWhether successfully extracted, wherein, the computing function of corr () expression normalizated correlation coefficient;

2.-8, make k=k+1, then returning step 2.-2 continues the digital 8bit gray level image of next bi-level digital watermark to be extracted is processed, until the digital 8bit gray level image of K bi-level digital watermark to be extracted all is disposed, wherein, "=" among the k=k+1 is assignment;

2.-9, be transferred to K the bi-level digital watermark W that end is extracted in digital watermarking from the digital watermarking built-in end ₁, W ₂..., W _k..., W _KIn select arbitrarily a bi-level digital watermark as Query Information, should optional bi-level digital watermark be designated as W _z, then calculate respectively W _zWith K bi-level digital watermark W ' that extracts ₁, W ' ₂..., W ' _k..., W ' _KNormalizated correlation coefficient, correspondence is designated as r ₁, r ₂..., r _k..., r _K, r ₁=corr (W _z, W ' ₁), r ₂=corr (W _z, W ' ₂), r _k=corr (W _z, W ' _k), r _K=corr (W _z, W ' _k), at last according to K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KReturn the digital 8bit gray level image that the corresponding K width of cloth embeds the bi-level digital watermark by large and little order, realize the retrieval of digital picture, wherein, 1≤z≤K, W ' ₁The 1st the bi-level digital watermark that expression extracts, W ' ₂The 2nd the bi-level digital watermark that expression extracts, W ' _kK the bi-level digital watermark that expression extracts, W ' _kK the bi-level digital watermark that expression extracts.

Described step 2.-5 in the order

Expression TFA _kMiddle coordinate position is

The coefficient value of coefficient, order

Expression TFAD _kMiddle coordinate position is

The coefficient value of coefficient, order

Expression TFAI _kMiddle coordinate position is

The coefficient value of coefficient, then have

Wherein, symbol

The maximum integer less than himself is got in expression, $1\≤\widetilde{i_k^{\′}}\≤(I_k^{\′}/2^{{L_k}^{\′}}),$ $1\≤\widetilde{j_k^{\′}}\≤(J_k^{\′}/2^{{L_k}^{\′}});$

Described step detailed process 2.-6 is: c1, with TFAD _kIn the current Coefficient Definition of processing be current coefficient, with Lg _kIn the current element definition of processing be currentElement; C2, suppose current coefficient coordinate position for (m ' _k, n ' _k), then the coefficient value with current coefficient is designated as tfad _k(m ' _k, n ' _k), the coordinate position of supposing currentElement for (m ' _k, n ' _k), then the logical value with currentElement is designated as lg _k(m ' _k, n ' _k), wherein, 1≤m ' _k≤ M ' _k, 1≤n ' _k≤ N ' _kC3, according to tfad _k(m ' _k, n ' _k) and lg _k(m ' _k, n ' _k), extract bi-level digital watermark W ' _kMiddle coordinate position be (m ' _k, n ' _k) the pixel value of pixel, be designated as w ' _k(m ' _k, n ' _k), specific as follows: as to work as lg _k(m ' _k, n ' _k)=0 and tfad _k(m ' _k, n ' _k) ∈ [0,0.5) time, w ' put _k(m ' _k, n ' _k)=0; Work as lg _k(m ' _k, n ' _k)=0 and tfad _k(m ' _k, n ' _k) ∈ [0.5,1) time, w ' put _k(m ' _k, n ' _k)=1; Work as lg _k(m ' _k, n ' _k)=1 and tfad _k(m ' _k, n ' _k) ∈ [0,0.5) time, w ' put _k(m ' _k, n ' _k)=1; Work as lg _k(m ' _k, n ' _k)=1 and tfad _k(m ' _k, n ' _k) ∈ [0.5,1) time, w ' put _k(m ' _k, n ' _k)=0; C4, with TFAD _kThe pending coefficient of the middle next one is as current coefficient, with Lg _kThen the pending element of the middle next one returns step c2 and continues to carry out, until TFAD as currentElement _kIn all coefficients and Lg _kIn all elements be disposed, realize bi-level digital watermark W ' _kExtraction.

Described step is determined the bi-level digital watermark W that is used for copyright protection that the digital watermarking built-in end embeds according to the size of r in 2.-7 _kThe detailed process of whether successfully being extracted is: if the value of r is 1, then determine the bi-level digital watermark w ' that extracts _kBi-level digital watermark W with the embedding of digital watermarking built-in end _kIn full accord, show the bi-level digital watermark W that the digital watermarking built-in end embeds _kNondestructively extracted, be can be used for copyright protection; If the value of r is between 1～δ _TBetween, then determine the bi-level digital watermark W ' that extracts _kBi-level digital watermark W with the embedding of digital watermarking built-in end _kSimilitude very large, this moment can be from the bi-level digital watermark W ' that extracts _kIn recognize the bi-level digital watermark W that the digital watermarking built-in end embeds _k, show the bi-level digital watermark W that the digital watermarking built-in end embeds _kSuccessfully extracted, be can be used for copyright protection, wherein, δ _TBe the watermark extracting thresholding of setting; If the value of r is less than δ _T, then determine the bi-level digital watermark W ' that extracts _kBi-level digital watermark W with the embedding of digital watermarking built-in end _kSimilitude very little, this moment can't be from the bi-level digital watermark W ' that extracts _kIn recognize the bi-level digital watermark W that the digital watermarking built-in end embeds _k, show the bi-level digital watermark W that the digital watermarking built-in end embeds _kExtract unsuccessfully.

Described step 2.-9 in according to K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KReturning the detailed process that the corresponding K width of cloth embeds the digital 8bit gray level image that the bi-level digital watermark is arranged by large and little order is: d1, find out K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KA normalizated correlation coefficient of intermediate value maximum is designated as maxr ₁, suppose maxr ₁Be K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KIn the individual normalizated correlation coefficient of k ', then returning to embed has the individual bi-level digital watermark of k ' W _{K '}Digital 8bit gray level image, as the 1st width of cloth image that meets the user search requirement, wherein, 1≤k '≤K; D2, find out a normalizated correlation coefficient of remaining K-1 normalizated correlation coefficient intermediate value maximum, be designated as maxr ₂, suppose maxr ₂Be K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KIn k " individual normalizated correlation coefficient, then returning to embed has k " individual bi-level digital watermark W _{K "}Digital 8bit gray level image, as the 2nd width of cloth image that meets the user search requirement, wherein, 1≤k "≤K; D3, find out a normalizated correlation coefficient of remaining K-2 normalizated correlation coefficient intermediate value maximum, be designated as maxr ₃, suppose maxr ₂Be K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KIn k ' " individual normalizated correlation coefficient, then returning to embed has k ' " individual bi-level digital watermark W _{K ' "}Digital 8bit gray level image, as the 3rd width of cloth image that meets the user search requirement, wherein, 1≤k ' "≤K; D4, according to the operation of steps d 3, continue in an identical manner to find out a normalizated correlation coefficient of each remaining normalizated correlation coefficient intermediate value maximum, return the digital 8bit gray level image that corresponding embedding has the bi-level digital watermark with this, until remaining last normalizated correlation coefficient is designated as maxr with last normalizated correlation coefficient _K, suppose maxr _KBe K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KIn k* normalizated correlation coefficient, then returning to embed has k* bi-level digital watermark W _K*Digital 8bit gray level image, the last piece image that arrives as user search, wherein, 1≤k*≤K.

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

1) compares with the digital image search technology of existing text based and content; data waterprint embedded method of the present invention is incorporated into digital watermark technology in the image retrieval process; in digital picture works warehouse-in link searching object (i.e. the digital picture works of warehouse-in) is embedded the bi-level digital watermark; it is under the copyright protection; improve the fail safe of works in the digital image search process, protected copyright owner's interests.

2) compare with the digital image search technology of existing text based and content, digital watermarking extracting method of the present invention is incorporated into digital watermark technology in the image retrieval process, Query Information with the bi-level digital watermark that embeds during as digital image search, and with the bi-level digital watermark that extracts the characteristic information during as digital image search, guaranteed the specific aim of retrieval, avoid simultaneously the foundation of large capacity index data base, efficiently solved the problem of memory space.

3) compare with the digital image search technology of existing text based and content, because the bi-level digital watermark that embeds has good robustness, the robustness of watermark has guaranteed the robustness of the information that embeds, therefore after the digital picture works stand normal image processing or external attack, still can effectively extract the bi-level digital watermark information (characteristic information) of embedding, thereby guarantee the accuracy of image retrieval.

4) compare with the digital watermark technology of content integrity authentication for copyright protection with existing; digital watermarking embedding of the present invention and extracting method are incorporated into digital watermark technology in the image retrieval process; not only make digital watermarking keep traditional aspect functions such as copyright protection; and use it for Query Information and characteristic information in the image retrieval process, further expanded the range of application of digital watermark technology.

Description of drawings

Fig. 1 a is original Bird digital picture;

Fig. 1 b is original Aimi digital picture;

Fig. 1 c is original Flower digital picture;

Fig. 1 d is original Scenery digital picture;

Fig. 2 a is original bi-level digital watermark W ₁

Fig. 2 b is original bi-level digital watermark W ₂

Fig. 2 c is original bi-level digital watermark W ₃

Fig. 2 d is original bi-level digital watermark W ₄

Fig. 3 a is for embedding bi-level digital watermark W ₁After the Bird digital picture;

Fig. 3 b is for embedding bi-level digital watermark W ₂After the Aimi digital picture;

Fig. 3 c is for embedding bi-level digital watermark W ₃After the Flower digital picture;

Fig. 3 d is for embedding bi-level digital watermark W ₄After the Scenerv digital picture;

The bi-level digital watermark of Fig. 4 a for from the digital picture shown in Fig. 3 a, extracting;

The bi-level digital watermark of Fig. 4 b for from the digital picture shown in Fig. 3 b, extracting;

The bi-level digital watermark of Fig. 4 c for from the digital picture shown in Fig. 3 c, extracting;

The bi-level digital watermark of Fig. 4 d for from the digital picture shown in Fig. 3 d, extracting;

Fig. 5 a is Query Information (original bi-level digital watermark W ₁);

Fig. 5 b is the first width of cloth digital picture in the result for retrieval of retrieving take Fig. 5 a as Query Information;

Fig. 5 c is the second width of cloth digital picture in the result for retrieval of retrieving take Fig. 5 a as Query Information;

Fig. 5 d is the 3rd width of cloth digital picture in the result for retrieval of retrieving take Fig. 5 a as Query Information;

Fig. 5 e is the 4th width of cloth digital picture in the result for retrieval of retrieving take Fig. 5 a as Query Information;

Fig. 6 a is Query Information (original bi-level digital watermark W ₂);

Fig. 6 b is the first width of cloth digital picture in the result for retrieval of retrieving take Fig. 6 a as Query Information;

Fig. 6 c is the second width of cloth digital picture in the result for retrieval of retrieving take Fig. 6 a as Query Information;

Fig. 6 d is the 3rd width of cloth digital picture in the result for retrieval of retrieving take Fig. 6 a as Query Information;

Fig. 6 e is the 4th width of cloth digital picture in the result for retrieval of retrieving take Fig. 6 a as Query Information;

Fig. 7 a is Query Information (original bi-level digital watermark W ₃);

Fig. 7 b is the first width of cloth digital picture in the result for retrieval of retrieving take Fig. 7 a as Query Information;

Fig. 7 c is the second width of cloth digital picture in the result for retrieval of retrieving take Fig. 7 a as Query Information;

Fig. 7 d is the 3rd width of cloth digital picture in the result for retrieval of retrieving take Fig. 7 a as Query Information;

Fig. 7 e is the 4th width of cloth digital picture in the result for retrieval of retrieving take Fig. 7 a as Query Information;

Fig. 8 a is Query Information (original bi-level digital watermark W ₄);

Fig. 8 b is the first width of cloth digital picture in the result for retrieval of retrieving take Fig. 8 a as Query Information;

Fig. 8 c is the second width of cloth digital picture in the result for retrieval of retrieving take Fig. 8 a as Query Information;

Fig. 8 d is the 3rd width of cloth digital picture in the result for retrieval of retrieving take Fig. 8 a as Query Information;

Fig. 8 e is the 4th width of cloth digital picture in the result for retrieval of retrieving take Fig. 8 a as Query Information;

Fig. 9 a is the digital picture of Fig. 3 a behind [9 * 9] window medium filtering;

Fig. 9 b is the digital picture of Fig. 3 b behind [9 * 9] window medium filtering;

Fig. 9 c is the digital picture of Fig. 3 c behind [9 * 9] window medium filtering;

Fig. 9 d is the digital picture of Fig. 3 d behind [9 * 9] window medium filtering;

The bi-level digital watermark of Figure 10 a for from Fig. 9 a, extracting;

The bi-level digital watermark of Figure 10 b for from Fig. 9 b, extracting;

The bi-level digital watermark of Figure 10 c for from Fig. 9 c, extracting;

The bi-level digital watermark of Figure 10 d for from Fig. 9 d, extracting;

Figure 11 a is Query Information (original bi-level digital watermark W ₂);

Figure 11 b is the first width of cloth digital picture in the result for retrieval of retrieving take Figure 11 a as Query Information;

Figure 11 c is the second width of cloth digital picture in the result for retrieval of retrieving take Figure 11 a as Query Information;

Figure 11 d is the 3rd width of cloth digital picture in the result for retrieval of retrieving take Figure 11 a as Query Information;

Figure 11 e is the 4th width of cloth digital picture in the result for retrieval of retrieving take Figure 11 a as Query Information;

Figure 12 a is the digital picture of Fig. 3 a after JPEG compression (8% quality factor);

Figure 12 b is the digital picture of Fig. 3 b after JPEG compression (8% quality factor);

Figure 12 c is the digital picture of Fig. 3 c after JPEG compression (8% quality factor);

Figure 12 d is the digital picture of Fig. 3 d after JPEG compression (8% quality factor);

The bi-level digital watermark of Figure 13 a for from Figure 12 a, extracting;

The bi-level digital watermark of Figure 13 b for from Figure 12 b, extracting;

The bi-level digital watermark of Figure 13 c for from Figure 12 c, extracting;

The bi-level digital watermark of Figure 13 d for from Figure 12 d, extracting;

Figure 14 a is Query Information (original bi-level digital watermark W ₃);

Figure 14 b is the first width of cloth digital picture in the result for retrieval of retrieving take Figure 14 a as Query Information;

Figure 14 c is the second width of cloth digital picture in the result for retrieval of retrieving take Figure 14 a as Query Information;

Figure 14 d is the 3rd width of cloth digital picture in the result for retrieval of retrieving take Figure 14 a as Query Information;

Figure 14 e is the 4th width of cloth digital picture in the result for retrieval of retrieving take Figure 14 a as Query Information;

Figure 15 a is Fig. 3 a through the stack average is digital picture after 0 variance is 0.05 Gaussian noise;

Figure 15 b is Fig. 3 b through the stack average is digital picture after 0 variance is 0.05 Gaussian noise;

Figure 15 c is Fig. 3 c through the stack average is digital picture after 0 variance is 0.05 Gaussian noise;

Figure 15 d is Fig. 3 d through the stack average is digital picture after 0 variance is 0.05 Gaussian noise;

The bi-level digital watermark of Figure 16 a for from Figure 15 a, extracting;

The bi-level digital watermark of Figure 16 b for from Figure 15 b, extracting;

The bi-level digital watermark of Figure 16 c for from Figure 15 c, extracting;

The bi-level digital watermark of Figure 16 d for from Figure 15 d, extracting;

Figure 17 a is Query Information (original bi-level digital watermark W ₄);

Figure 17 b is the first width of cloth digital picture in the result for retrieval of retrieving take Figure 17 a as Query Information;

Figure 17 c is the second width of cloth digital picture in the result for retrieval of retrieving take Figure 17 a as Query Information;

Figure 17 d is the 3rd width of cloth digital picture in the result for retrieval of retrieving take Figure 17 a as Query Information;

Figure 17 e is the 4th width of cloth digital picture in the result for retrieval of retrieving take Figure 17 a as Query Information;

Figure 18 a is the digital picture of Fig. 3 a after cutting the upper right corner 256 * 256 pixels;

Figure 18 b is the digital picture of Fig. 3 b after cutting the upper right corner 256 * 256 pixels;

Figure 18 c is the digital picture of Fig. 3 c after cutting the upper right corner 256 * 256 pixels;

Figure 18 d is the digital picture of Fig. 3 d after cutting the upper right corner 256 * 256 pixels;

The bi-level digital watermark of Figure 19 a for from Figure 18 a, extracting;

The bi-level digital watermark of Figure 19 b for from Figure 18 b, extracting;

The bi-level digital watermark of Figure 19 c for from Figure 18 c, extracting;

The bi-level digital watermark of Figure 19 d for from Figure 18 d, extracting;

Figure 20 a is Query Information (original bi-level digital watermark W ₁);

Figure 20 b is the first width of cloth digital picture in the result for retrieval of retrieving take Figure 20 a as Query Information;

Figure 20 c is the second width of cloth digital picture in the result for retrieval of retrieving take Figure 20 a as Query Information;

Figure 20 d is the 3rd width of cloth digital picture in the result for retrieval of retrieving take Figure 20 a as Query Information;

Figure 20 e is the 4th width of cloth digital picture in the result for retrieval of retrieving take Figure 20 a as Query Information.

Embodiment

Embodiment is described in further detail the present invention below in conjunction with accompanying drawing.

Larger for required memory space in the existing digital image search technology; the limitation that fail safe is relatively poor; the present invention proposes a kind of digital watermarking embedding and extracting method for image retrieval and copyright protection; it is digital watermark technology to be incorporated in the digital image search process that the present invention embeds in (image works warehouse-in) process in digital watermarking; specifically (sign such as the author by the structure bianry image relevant with the original digital image content; copyright marking; sequence number; date or icon etc.) watermark is embedded in the original digital image works as bi-level digital; the Query Information during again as digital image search in digital watermarking leaching process (image works retrieval link) simultaneously; and the bi-level digital watermark of extracting is used on the one hand the characteristic information of digital image search process; reach quick by the degree of correlation between calculating and the Query Information; digital image search purpose accurately; the copyright information that is used on the other hand the digital image security field; play the purposes such as digital image copyright protection, so effectively improved the fail safe of digital picture works.Digital watermarking embedding of the present invention and extracting method have not only been avoided the foundation of large capacity aspect indexing database, efficiently solve the problem of memory space, and after standing normal image processing and external attack, also can retrieve quickly and accurately required digital picture.Further, in retrieving, digital picture is under the copyright protection all the time, has improved the fail safe of works, has protected copyright owner's interests.

A kind of data waterprint embedded method for image retrieval and copyright protection that the present invention proposes, its processing procedure is: at first, the original figure 8bit gray level image of bi-level digital watermark to be embedded is carried out normalized; Secondly, the digital picture that obtains after the normalized is carried out two-dimensional discrete wavelet conversion, obtain first a small echo ll channel and a plurality of the first wavelet details subgraph; The first small echo ll channel is decomposed into a fractional part the first small echo ll channel and integer part first a small echo ll channel; Again, pixel value according to each pixel in the bi-level digital watermark to be embedded, coefficient value scope in conjunction with each coefficient in fractional part the first small echo ll channel, coefficient value to each coefficient in fractional part the first small echo ll channel carries out the two-value quantification treatment, realize the embedding of bi-level digital watermark, and return a two-valued function table; Then, new fractional part the first small echo ll channel and integer part the first small echo ll channel that obtains after the two-value quantification treatment is combined into the first new small echo ll channel; The first new small echo ll channel and a plurality of the first wavelet details subgraph are carried out the 2-d discrete wavelet inverse transformation together, obtain embedding the digital picture of bi-level digital watermark; At last, the bi-level digital watermark is reached the logical table that in the bi-level digital watermark embed process, returns and be transferred to digital watermarking extraction end.

Data waterprint embedded method of the present invention, it specifically may further comprise the steps:

1.-1, at the digital watermarking built-in end, suppose to have K bi-level digital watermark to be embedded and the original figure 8bit gray level image of K bi-level digital watermark to be embedded, k bi-level digital watermark is designated as W _k, W _k={ w _k| w _k(m _k, n _k)=0 or 1,1≤m _k≤ M _k, 1≤n _k≤ N _k, the original figure 8bit gray level image of k bi-level digital watermark to be embedded is designated as F _k, F _k={ f _k| 0≤f _k(i _k, j _k)≤255,1≤i _k≤ I _k, 1≤j _k≤ J _k, wherein, 1≤k≤K, K 〉=1, W _kResolution be M _k* N _k, w _k(m _k, n _k) expression W _kMiddle coordinate position is (m _k, n _k) the pixel value of pixel, F _kResolution be I _k* J _k, f _k(i _k, j _k) expression F _kMiddle coordinate position is (i _k, j _k) the pixel value of pixel.Be without loss of generality, for processing conveniently, suppose k bi-level digital watermark W at this _kVertical resolution M _kWith horizontal resolution N _kEquate, i.e. M _k=N _k, suppose simultaneously the original figure 8bit gray level image F of k bi-level digital watermark to be embedded _kVertical resolution I _kWith horizontal resolution J _kEquate, i.e. I _k=J _k, and satisfy

$J_k/N_k=2^{L_k},$ L _k≥1。

At this; K bi-level digital watermark to be embedded is bianry image; be selected from respectively and the original digital image content is relevant separately bianry image; can be image author signature, icon, copyright marking, sequence number and date created etc.; some content characteristic that had both comprised original digital image; Query Information when can be used for image retrieval has reflected again the copyright information of original digital image to can be used for the purposes such as image copyright protection.

At this, the original digital image of bi-level digital watermark to be embedded is all the 8bit gray level image in the image library.

1.-2 the original figure 8bit gray level image of, supposing the current bi-level digital watermark to be embedded of processing is the original figure 8bit gray level image F of k bi-level digital watermark to be embedded _k

1.-3, to F _kCarry out normalized, with F _kThe digital picture F ' that after normalized, obtains _kMiddle coordinate position is (i _k, j _k) the pixel value of pixel be designated as f ' _k(i _k, j _k), f ' _k(i _k, j _k)=f _k(i _k, j _k)/255, wherein, 1≤i _k≤ I _k, 1≤j _k≤ J _k

1.-4, to F ' _kCarry out L _kThe level two-dimensional discrete wavelet conversion obtains first a small echo ll channel and a plurality of the first wavelet details subgraph, with F ' _kThe first corresponding small echo ll channel is designated as FA _k, wherein, L _k=log ₂(I _k/ M _k)=log ₂(J _k/ N _k), FA _kResolution be

1.-5, with FA _kBe decomposed into fractional part and integer part two parts data subgraph, obtain a fractional part the first small echo ll channel and integer part first a small echo ll channel, be designated as respectively FAD _kAnd FAI _k, in this hypothesis order

Expression FA _kMiddle coordinate position is

The coefficient value of coefficient, order

Expression FAD _kMiddle coordinate position is

The coefficient value of coefficient, order

Expression FAI _kMiddle coordinate position is

The coefficient value of coefficient, then have

Wherein, symbol

The maximum integer less than himself is got in expression,

1.-6, according to W _kIn the pixel value of each pixel, to FAD _kIn the coefficient value of each coefficient carry out the two-value quantification treatment, realize bi-level digital watermark W _kEmbedding, and return a two-valued function table, the fractional part first small echo ll channel that will obtain after the two-value quantification treatment is designated as FAD ' _k, be Lg with the two-valued function souvenir that returns _k

In this specific embodiment, step detailed process 1.-6 is: a1, with FAD _kIn the current Coefficient Definition of processing be current coefficient; A2, suppose that the coordinate position of current coefficient is (m _k, n _k), then the coefficient value with current coefficient is designated as fad _k(m _k, n _k), wherein, 1≤m _k≤ M _k, 1≤n _k≤ N _kA3, according to fad _k(m _k, n _k) and W _kMiddle coordinate position is (m _k, n _k) the pixel value w of pixel _k(m _k, n _k), calculate FAD ' _kMiddle coordinate position is (m _k, n _k) the coefficient value of coefficient, be designated as fad ' _k(m _k, n _k), and return two-valued function table Lg _kMiddle coordinate position is (m _k, n _k) the logical value lg of element _k(m _k, n _k), specific as follows: as to work as w _k(m _k, n _k)=0 and fad _k(m _k, n _k) ∈ [0,0.5) time, with fad _k(m _k, n _k) be quantified as 0.25, i.e. fad ' _k(m _k, n _k)=0.25, lg _k(m _k, n _k)=0; Work as w _k(m _k, n _k)=0 and fad _k(m _k, n _k) ∈ [0.5,1), with fad _k(m _k, n _k) be quantified as 0.75, i.e. fad ' _k(m _k, n _k)=0.75, lg _k(m _k, n _k)=1; Work as w _k(m _k, n _k)=1 and fad _k(m _k, n _k) ∈ [0,0.5), with fad _k(m _k, n _k) be quantified as 0.25, i.e. fad ' _k(m _k, n _k)=0.25, lg _k(m _k, n _kW is worked as in)=1 _k(m _k, n _k)=1 and fad _k(m _k, n _k) ∈ [0.5,1), with fad _k(m _k, n _k) be quantified as 0.75, i.e. fad ' _k(m _k, n _k)=0.75, lg _k(m _k, n _k)=0; A4, with FAD _kThen the pending coefficient of the middle next one returns step a2 and continues to carry out, until FAD as current coefficient _kIn all coefficient processing complete, realize bi-level digital watermark W _kEmbedding, obtain fractional part the first small echo ll channel FAD ' that after the two-value quantification treatment, obtains _kIn actual process, if FAD _kResolution and W _kResolution inconsistent, namely both images are not of uniform size causes, then can be first to W _kCarry out preliminary treatment and make its resolution and FAD _kResolution consistent, carry out again bi-level digital watermark W _kEmbedding.

1.-7, with FAD ' _kAnd FAI _kBe combined into F ' _kCorresponding the first new small echo ll channel is designated as FA ' _k

In this specific embodiment, step detailed process 1.-7 is: b1, with FAD ' _kIn the current Coefficient Definition of processing be current the first coefficient, with FAI _kIn the current Coefficient Definition of processing be current the second coefficient; B2, suppose that the coordinate position of current the first coefficient is

Then the coefficient value with current the first coefficient is designated as

The coordinate position of supposing current the second coefficient is

Then the coefficient value with current the second coefficient is designated as

Wherein,

B3, basis With

Calculate FA ' _kMiddle coordinate position is

The coefficient value of coefficient, be designated as

B4, with FAD ' _kThe pending coefficient of the middle next one is as current the first coefficient, with FAI _kThen the pending coefficient of the middle next one returns step b2 and continues to carry out, until FAD ' as current the second coefficient _kAnd FAI _kIn all coefficient processing complete, obtain F ' _kCorresponding the first new small echo ll channel FA ' _k

1.-8, with FA ' _kAnd F ' _kCorresponding a plurality of the first wavelet details subgraphs carry out L together again _kThe inverse transformation of level 2-d discrete wavelet, obtaining embedding has bi-level digital watermark W _kDigital picture, be designated as F " _k, finished thus bi-level digital watermark W _kEmbedding be original figure 8bit gray level image works F _kThe warehouse-in process.

1.-9, make k=k+1, then returning step 1.-2 continues the original figure 8bit gray level image of next bi-level digital watermark to be embedded is processed, until the original figure 8bit gray level image of K bi-level digital watermark to be embedded all is disposed, wherein, "=" among the k=k+1 is assignment.

1.-10, with K bi-level digital watermark W ₁, W ₂..., W _k..., W _KAnd the two-valued function table Lg that in the bi-level digital watermark embed process, returns ₁, Lg ₂..., Lg _k..., Lg _KBe transferred to digital watermarking and extract end, wherein, W ₁Represent the 1st bi-level digital watermark, W ₂Represent the 2nd bi-level digital watermark, W _KRepresent K bi-level digital watermark, Lg ₁Be illustrated in bi-level digital watermark W ₁The two-valued function table that returns in the telescopiny, Lg ₂Be illustrated in bi-level digital watermark W ₂The two-valued function table that returns in the telescopiny, Lg _KBe illustrated in bi-level digital watermark W _KThe two-valued function table that returns in the telescopiny.

A kind of digital watermarking extracting method for image retrieval and copyright protection that the present invention proposes, its processing procedure is: at first, the digital 8bit gray level image of bi-level digital watermark to be extracted is carried out normalized; Secondly, the digital picture that obtains after the normalized is carried out two-dimensional discrete wavelet conversion, obtain second a small echo ll channel and a plurality of the second wavelet details subgraph; The second small echo ll channel is decomposed into a fractional part the second small echo ll channel and integer part second a small echo ll channel; Again, be transferred to the two-valued function table that end is extracted in digital watermarking according to the digital watermarking built-in end, and the coefficient value scope of each coefficient in fractional part the second small echo ll channel, carry out two-value and differentiate, extract the bi-level digital watermark; Then, be transferred to the coefficient correlation that bi-level digital watermark with the bi-level digital watermark that extracts of end are extracted in digital watermarking by calculating the digital watermarking built-in end, determine whether the bi-level digital watermark that extracts is the bi-level digital watermark that is used for copyright protection that the digital watermarking built-in end embeds; At last, utilize coefficient correlation to realize the retrieval of digital picture.

Digital watermarking extracting method of the present invention, it specifically may further comprise the steps:

2.-1, extract end in digital watermarking, suppose to have the individual bi-level digital watermark to be extracted of K and the digital 8bit gray level image of K bi-level digital watermark to be extracted, k bi-level digital watermark is designated as W ' _k, W ' _k=w ' _k| w ' _k(m ' _k, n ' _k)=0 or 1,1≤m ' _k≤ M ' _k, 1≤n ' _k≤ N ' _k, the digital 8bit gray level image of k bi-level digital watermark to be extracted is designated as TF _k, TF _k={ tf _k| 0≤tf _k(i ' _k, j ' _k)≤255,1≤i ' _k≤ I ' _k, 1≤j ' _k≤ J ' _k, wherein, 1≤k≤K, K 〉=1, W ' _kResolution be M ' _k* N ' _k, at this hypothesis M ' _k=N ' _k, and W ' _kThe resolution of the bi-level digital watermark that embeds of resolution and digital watermarking built-in end identical, w ' _k(m ' _k, n ' _k) expression W ' _kMiddle coordinate position be (m ' _k, n ' _k) the pixel value of pixel, TF _kResolution be I ' _k* J ' _k, at this hypothesis I ' _k=J ' _k, and TF _kResolution identical with the resolution of the original figure 8bit gray level image of digital watermarking built-in end bi-level digital watermark to be embedded, tf _k(i ' _k, j ' _k) expression TF _kMiddle coordinate position be (i ' _k, j ' _k) the pixel value of pixel.

At this, the digital picture of bi-level digital watermark to be extracted in the image library (or image to be retrieved) is all the 8bit gray level image.

2.-2 the digital 8bit gray level image of, supposing the current bi-level digital watermark to be extracted of processing is the digital 8bit gray level image TF of k bi-level digital watermark to be extracted _k

2.-3, to TF _kCarry out normalized, with TF _kThe digital picture TF ' that after normalized, obtains _kMiddle coordinate position be (i ' _K,J ' _k) the pixel value of pixel be designated as tf ' _k(i ' _k, j ' _k), tf ' _k(i ' _k, j ' _k)=tf _k(i ' _k, j ' _k)/255, wherein, 1≤i ' _k≤ I ' _k, 1≤j ' _k≤ J ' _k

2.-4, to TF ' _kCarry out L ' _kThe level two-dimensional discrete wavelet conversion obtains second a small echo ll channel and a plurality of the second wavelet details subgraph, with TF ' _kThe second corresponding small echo ll channel is designated as TFA _k, wherein, L ' _k=log ₂(I ' _k/ M ' _k)=log ₂(J ' _k/ N ' _k), TFA _kResolution be

2.-5, with TFA _kBe decomposed into fractional part and integer part two parts data subgraph, obtain a fractional part the second small echo ll channel and integer part second a small echo ll channel, be designated as respectively TFAD _kAnd TFAI _k, in this hypothesis order

Expression TFA _kMiddle coordinate position is

The coefficient value of coefficient, order

Expression TFAD _kMiddle coordinate position is

The coefficient value of coefficient, order

Expression TFAI _kMiddle coordinate position is

The coefficient value of coefficient, then have

Wherein, symbol

The maximum integer less than himself is got in expression,

2.-6, be transferred to the two-valued function table Lg that end is extracted in digital watermarking according to the digital watermarking built-in end _kAnd TFAD _kIn the coefficient value of each coefficient, extract bi-level digital watermark W ' _k

In this specific embodiment, step detailed process 2.-6 is: c1, with TFAD _kIn the current Coefficient Definition of processing be current coefficient, with Lg _kIn the current element definition of processing be currentElement; C2, suppose current coefficient coordinate position for (m ' _k, n ' _k), then the coefficient value with current coefficient is designated as tfad _k(m ' _k, n ' _k), the coordinate position of supposing currentElement for (m ' _k, n ' _k), then the logical value with currentElement is designated as lg _k(m ' _k, n ' _k), wherein, 1≤m ' _k≤ M _k, 1≤n ' _k≤ N ' _kC3, according to tfad _k(m ' _k, n ' _k) and lg _k(m ' _k, n ' _k), extract bi-level digital watermark W ' _kMiddle coordinate position be (m ' _k, n ' _k) the pixel value of pixel, be designated as w ' _k(m ' _k, n ' _k), specific as follows: as to work as lg _k(m ' _k, n ' _k)=0 and tfad _k(m ' _k, n ' _k) ∈ [0,0.5) time, w ' put _k(m ' _k, n ' _k)=0; Work as lg _k(m ' _k, n ' _k)=0 and tfad _k(m ' _k, n ' _k) ∈ [0.5,1) time, w ' put _k(m ' _k, n ' _k)=1; Work as lg _k(m ' _k, n ' _k)=1 and tfad _k(m ' _k, n ' _k) ∈ [0,0.5) time, w ' put _k(m ' _k, n ' _k)=1; Work as lg _k(m ' _k, n ' _k)=1 and tfad _k(m ' _k, n ' _k) ∈ [0.5,1) time, w ' put _k(m ' _k, n ' _k)=0; C4, with TFAD _kThe pending coefficient of the middle next one is as current coefficient, with Lg _kThen the pending element of the middle next one returns step c2 and continues to carry out, until TFAD as currentElement _kIn all coefficients and Lg _kIn all elements be disposed, realize bi-level digital watermark W ' _kExtraction.

2.-7, calculate the digital watermarking built-in end and be transferred to the bi-level digital watermark W that end is extracted in digital watermarking _kWith the bi-level digital watermark W ' that extracts _kNormalizated correlation coefficient, be designated as r, r=corr (W _k, W ' _k), then determine the bi-level digital watermark W that is used for copyright protection that the digital watermarking built-in end embeds according to the size of r _kWhether successfully extracted, the bi-level digital watermark that extracts can be used for the purposes such as copyright protection of each original digital image, wherein, and the computing function of corr () expression normalizated correlation coefficient.

In this specific embodiment, step is determined the bi-level digital watermark W that is used for copyright protection that the digital watermarking built-in end embeds according to the size of r in 2.-7 _kThe detailed process of whether successfully being extracted is: if the value of r is 1, then determine the bi-level digital watermark W ' that extracts _kBi-level digital watermark W with the embedding of digital watermarking built-in end _kIn full accord, show the bi-level digital watermark W that the digital watermarking built-in end embeds _kNondestructively extracted, be can be used for copyright protection.If the value of r is between 1～δ _TBetween, the bi-level digital watermark W ' that extracts then is described _kBi-level digital watermark W with the built-in end embedding _kThere is some difference, but have very large similitude between the two, the bi-level digital watermark W ' that can namely extract this moment from extracting the result _kIn recognize the bi-level digital watermark W that the digital watermarking built-in end embeds _k, show the bi-level digital watermark W that the digital watermarking built-in end embeds _kSuccessfully extracted, be can be used for copyright protection.The value of normalizated correlation coefficient shows the bi-level digital watermark W ' that extracts more close to 1 _kBi-level digital watermark W with the built-in end embedding _kMore similar, more easily recognize the bi-level digital watermark W of embedding _k, extraction effect is better.Wherein, δ _TBe the watermark extracting thresholding, but general value is 0.5.If the value of r is less than δ _T, the bi-level digital watermark W ' that extracts then is described _kBi-level digital watermark W with the built-in end embedding _kCorrelation is very little, the bi-level digital watermark W ' that at this moment can't namely extract from extracting the result _kIn recognize the bi-level digital watermark W of embedding _k, extract unsuccessfully.

2.-8, make k=k+1, then returning step 2.-2 continues the digital 8bit gray level image of next bi-level digital watermark to be extracted is processed, until the digital 8bit gray level image of K bi-level digital watermark to be extracted all is disposed, wherein, "=" among the k=k+1 is assignment.

2.-9, be transferred to K the bi-level digital watermark W that end is extracted in digital watermarking from the digital watermarking built-in end ₁, W ₂..., W _k..., W _KIn select arbitrarily a bi-level digital watermark as Query Information, should optional bi-level digital watermark be designated as W _z, then calculate respectively W _zWith K bi-level digital watermark W ' that extracts ₁, W ' ₂..., W ' _k..., W ' _kNormalizated correlation coefficient, correspondence is designated as r ₁, r ₂..., r _k..., r _K, r ₁=corr (W _z, W ' ₁), r ₂=corr (W _z, W ' ₂), r _k=corr (W _z, W ' _k), r _K=corr (W _z, W ' _K), at last according to K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KReturn the digital 8bit gray level image that the corresponding K width of cloth embeds the bi-level digital watermark by large and little order, realize the retrieval of digital picture, wherein, 1≤z≤K, W ' ₁The 1st the bi-level digital watermark that expression extracts, W ' ₂The 2nd the bi-level digital watermark that expression extracts, W ' _kK the bi-level digital watermark that expression extracts, W ' _KK the bi-level digital watermark that expression extracts.

In this specific embodiment, step 2.-9 in according to K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KReturning corresponding K width of cloth embedding by large and little order has the detailed process of the digital 8bit gray level image of bi-level digital watermark to be: d1, this K of comparison normalizated correlation coefficient r ₁, r ₂..., r _k..., r _K, the normalizated correlation coefficient that the value of finding out is maximum is designated as maxr ₁, suppose maxr ₁Be K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KIn the individual normalizated correlation coefficient of k ', then returning to embed has the individual bi-level digital watermark of k ' W _{K '}Digital 8bit gray level image, as the 1st width of cloth image that meets the user search requirement, wherein, 1≤k '≤K; D2, for K-1 remaining normalizated correlation coefficient, continue relatively their size, a normalizated correlation coefficient of the value of finding out maximum is designated as maxr ₂, suppose maxr ₂Be K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KIn k " individual normalizated correlation coefficient, then returning to embed has k " individual bi-level digital watermark W _{K "}Digital 8bit gray level image, as the 2nd width of cloth image that meets the user search requirement, wherein, 1≤k "≤K; D3, find out a normalizated correlation coefficient of remaining K-2 normalizated correlation coefficient intermediate value maximum, be designated as maxr ₂, suppose maxr ₂Be K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KIn k ' " individual normalizated correlation coefficient, then returning to embed has k ' " individual bi-level digital watermark W _{K ' "}Digital 8bit gray level image, as the 3rd width of cloth image that meets the user search requirement, wherein, 1≤k ' "≤K; D4, by that analogy, namely according to the operation of steps d 3, continue in an identical manner to find out a normalizated correlation coefficient of each remaining normalizated correlation coefficient intermediate value maximum, return the digital 8bit gray level image that corresponding embedding has the bi-level digital watermark with this, until remaining last normalizated correlation coefficient is designated as maxr with last normalizated correlation coefficient _K, suppose maxr _KBe K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KIn k* normalizated correlation coefficient, then returning to embed has k* bi-level digital watermark W _K*Digital 8bit gray level image, as the last piece image that user search arrives, be arranged in the last of user search result, wherein, 1≤k*≤K.

For illustrating that better the present invention is used for the digital watermarking embedding of image retrieval and copyright protection and feasibility and the validity of extracting method, 4 width of cloth digital pictures in the image library are carried out experimental verification as example.

Experimental verification is carried out at the Matlab7.5 platform, the original digital image of bi-level digital watermark to be embedded is selected the gray level image of 8bit: Bird, Aimi, Flower and Scenery, resolution is all 512 * 512, respectively shown in Fig. 1 a, Fig. 1 b, Fig. 1 c and Fig. 1 d.Bi-level digital watermark W to be embedded ₁, W ₂, W ₃And W ₄Elect bianry image as, resolution is all 32 * 32, respectively shown in Fig. 2 a, Fig. 2 b, Fig. 2 c and Fig. 2 d.

The objective evaluation of digital picture quality adopts Y-PSNR (PSNR) after the watermark of embedding bi-level digital, $\mathrm{PSNR}=-10\×\log 10\left(\frac{\underset{i=1}{\overset{I}{\mathrm{\Σ}}}\underset{j=1}{\overset{J}{\mathrm{\Σ}}}{(f(i,j)-f^{\′}(i,j))}^2}{I\×J\×f_{\max }^2}\right),$ Wherein, coordinate position is (i among the original digital image F of f (i, j) expression bi-level digital watermark to be embedded, the pixel value of pixel j), the middle coordinate position of digital picture F ' that f ' (i, j) expression embeds after the bi-level digital watermark is (i, the pixel value of pixel j), f _MaxThe max pixel value of expression original digital image F, the resolution of the digital picture F ' after I * J represents the original digital image F of bi-level digital watermark to be embedded and embeds the bi-level digital watermark.

Bi-level digital watermark extracting result's objective evaluation adopts normalizated correlation coefficient (r): $r(W,W^{\′})=\frac{\underset{m=1}{\overset{M}{\mathrm{\Σ}}}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}((w(m,n)-\stackrel{\‾}{w})\×(w^{\′}(m,n)-\stackrel{\‾}{w^{\′}})}{\sqrt{\underset{m=1}{\overset{M}{\mathrm{\Σ}}}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{(w(m,n)-\stackrel{\‾}{w})}^2\×\underset{m=1}{\overset{M}{\mathrm{\Σ}}}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}{(w^{\′}(m,n)-\stackrel{\‾}{w^{\′}})}^2}},$ Wherein, w (m, n) represents that coordinate position is the pixel value of the pixel of (m, n) among the original bi-level digital watermark W, and the middle coordinate position of bi-level digital watermark W ' that w ' (m, n) expression extracts is the pixel value of the pixel of (m, n),

With

The average that represents respectively all pixels among W and the W ', M represents the horizontal resolution of W and W ', N represents the vertical resolution of W and W '.In copyright protection; can determine whether the bi-level digital watermark that extracts is the bi-level digital watermark that the digital watermarking built-in end embeds according to the size of coefficient correlation; and in digital image search, can arrange according to the size of coefficient correlation the order of returning of result for retrieval.

Fig. 3 a embeds the bi-level digital watermark W shown in Fig. 2 a in the original Bird digital picture shown in Fig. 1 a ₁After watermark Bird digital picture, Fig. 3 b embeds the bi-level digital watermark W shown in Fig. 2 b in the original Aimi digital picture shown in Fig. 1 b ₂After watermark Aimi digital picture, Fig. 3 c embeds the bi-level digital watermark W shown in Fig. 2 c in the original Flower digital picture shown in Fig. 1 c ₃After watermark Flower digital picture, Fig. 3 d embeds the bi-level digital watermark W shown in Fig. 2 d in the original Scenery digital picture shown in Fig. 1 d ₄After watermark Scenery digital picture.By Fig. 3 a, Fig. 3 b, Fig. 3 c and Fig. 3 d as seen, than each original digital image of Fig. 1 a, Fig. 1 b, Fig. 1 c and Fig. 1 d, master, the objective quality of the watermark digit image after the watermark of embedding bi-level digital are all ideal.Subjective, human eye vision does not observe any variation, and objectively, Y-PSNR (PSNR) is all very high, has reached respectively 40.25dB, 41.23dB, 41.43dB and 40.76dB.Illustrate that the data waterprint embedded method that the present invention proposes is feasible, satisfied the requirement of watermark imperceptible fully, therefore can not have influence on the subsequent applications of original digital image.

The bi-level digital watermark W ' of Fig. 4 a for from Fig. 3 a watermark Bird digital picture, extracting ₁, the bi-level digital watermark W ' of Fig. 4 b for from Fig. 3 b watermark Aimi digital picture, extracting ₂, the bi-level digital watermark W ' of Fig. 4 c for from Fig. 3 c watermark Flower digital picture, extracting ₃, the bi-level digital watermark W ' of Fig. 4 d for from Fig. 3 d watermark Scenery digital picture, extracting ₄From Fig. 4 a, Fig. 4 b, Fig. 4 c and Fig. 4 d, can find out, be not affected or during external attack at the watermark digit image shown in Fig. 3 a, Fig. 3 b, Fig. 3 c and Fig. 3 d, can fully nondestructively extract respectively the bi-level digital watermark of embedding from the watermark digit image, original bi-level digital watermark all is 1 with the normalizated correlation coefficient r of the bi-level digital watermark that extracts.Illustrate that the digital watermarking extracting method that the present invention proposes is feasible, the digital watermarking that extracts can be used for the separately aspect purposes such as copyright protection of digital picture fully.

Fig. 5 a is with original bi-level digital watermark W ₁As Query Information, Fig. 5 b, Fig. 5 c, Fig. 5 d and Fig. 5 e are with original bi-level digital watermark W ₁The result for retrieval that returns during as Query Information, Fig. 6 a are with original bi-level digital watermark W ₂As Query Information, Fig. 6 b, Fig. 6 c, Fig. 6 d and Fig. 6 e are with original bi-level digital watermark W ₂The result for retrieval that returns during as Query Information, Fig. 7 a are with original bi-level digital watermark W ₃As Query Information, Fig. 7 b, Fig. 7 c, Fig. 7 d and Fig. 7 e are with original bi-level digital watermark W ₃The result for retrieval that returns during as Query Information, Fig. 8 a are with original bi-level digital watermark W ₄As Query Information, Fig. 8 b, Fig. 8 c, Fig. 8 d and Fig. 8 e are with original bi-level digital watermark W ₄The result for retrieval that returns during as Query Information.Retrieval is that the bi-level digital watermark in the digital picture to be retrieved is as characteristic information in the image library by extracting, and then the correlation of comparison query information and characteristic information is returned by large and little order according to coefficient correlation.Can find out from Fig. 5 a to Fig. 8 e, what result for retrieval had disclosed that the present invention proposes carries out feasibility and the specific aim of image retrieval based on digital watermark technology, can retrieve quickly and accurately the required digital picture of user.For example, the 1st width of cloth digital picture returned of the retrieval Query Result that needs most of user just each time, its normalizated correlation coefficient is a few width of cloth digital pictures much larger than the back.

The below carries out various attacks to the digital 8bit gray level image of bi-level digital watermark to be extracted in the image library and processes, and that verifies that the present invention proposes is used for image retrieval and the digital watermarking embedding of copyright protection and feasibility and the robustness of extracting method.

(1) medium filtering

Each width of cloth watermark digit image shown in Fig. 3 a, Fig. 3 b, Fig. 3 c and Fig. 3 d is carried out medium filtering to be processed, the filtering window size is chosen as [9 * 9], obtains filtered watermark Bird, Aimi, Flower and Scenery digital picture respectively shown in Fig. 9 a, Fig. 9 b, Fig. 9 c and Fig. 9 d.The bi-level digital watermark W ' of Figure 10 a for from the watermark Bird digital picture shown in Fig. 9 a, extracting ₁, the bi-level digital watermark W ' of Figure 10 b for from the watermark Aimi digital picture shown in Fig. 9 b, extracting ₂, the bi-level digital watermark W ' of Figure 10 c for from the watermark Flower digital picture shown in Fig. 9 c, extracting ₃, the bi-level digital watermark W ' of Figure 10 d for from the watermark Scenery digital picture shown in Fig. 9 d, extracting ₄Medium filtering for [9 * 9] large window is processed; with respect to Fig. 3 a; Fig. 3 b; original watermark digital picture shown in Fig. 3 c and Fig. 3 d; Fig. 9 a; Fig. 9 b; watermark Bird shown in Fig. 9 c and Fig. 9 d; Aimi; the detailed information of Flower and Scenery digital picture is fuzzyyer; Y-PSNR PSNR drops to respectively 30.22dB; 29.97dB; 28.76dB and 24.58dB; but by the watermark extracting result as can be known; the bi-level digital watermark that the present invention embeds has more satisfactory anti-filtering disposal ability; normalizated correlation coefficient r has reached respectively 0.941; 0.853; 0.854 with 0.641, can play the copyright protection effect.

With the original bi-level digital watermark W shown in Figure 11 a ₂As Query Information, watermark Bird, Aimi, Flower and Scenery digital picture shown in Fig. 9 a, Fig. 9 b, Fig. 9 c and Fig. 9 d to be retrieved, the result for retrieval that returns is shown in Figure 11 b, Figure 11 c, Figure 11 d and Figure 11 e.As seen, the watermark digit image although deteriroation of image quality is obvious, but still can retrieve user's results needed quickly and accurately in the image library after excessive window medium filtering is processed.

(2) JPEG lossy compression method

Watermark digit image shown in Fig. 3 a, Fig. 3 b, Fig. 3 c and Fig. 3 d is carried out the JPEG lossy compression method to be processed, the compression quality factor is chosen as 8%, and watermark Bird, the Aimi that obtains, Flower and Scenery digital picture are shown in Figure 12 a, Figure 12 b, Figure 12 c and Figure 12 d.The bi-level digital watermark W ' of Figure 13 a for from the watermark Bird digital picture shown in Figure 12 a, extracting ₁, the bi-level digital watermark W ' of Figure 13 b for from the watermark Aimi digital picture shown in Figure 12 b, extracting ₂, the bi-level digital watermark W ' of Figure 13 c for from the watermark Flower digital picture shown in Figure 12 c, extracting ₃, the bi-level digital watermark W ' of Figure 13 d for from the watermark Scenery digital picture shown in Figure 12 d, extracting ₄Be that 8% JPEG lossy compression method is processed for the compression quality factor; Figure 12 a; Figure 12 b; watermark Bird shown in Figure 12 c and Figure 12 d; Aimi; Flower and Scenery digital picture all present apparent in view blocking artifact; degeneration has occured in visual quality; Y-PSNR PSNR drops to respectively 31.97dB; 31.17dB; 31.09dB and 28.91dB; but by the watermark extracting result as can be known; the bi-level digital watermark that the present invention embeds has very desirable anti-JPEG lossy compression method disposal ability; normalizated correlation coefficient r has reached respectively 0.885; 0.866; 0.653 with 0.920, can play the copyright protection effect.

With the original bi-level digital watermark W shown in Figure 14 a ₃As Query Information, watermark Bird, Aimi, Flower and Scenery digital picture shown in Figure 12 a, Figure 12 b, Figure 12 c and Figure 12 d to be retrieved, the result for retrieval that returns is shown in Figure 14 b, Figure 14 c, Figure 14 d and Figure 14 e.As seen, after the watermark digit image is processed through the JPEG lossy compression method in the image library, although deteriroation of image quality is obvious, but still can retrieve quickly and accurately user's results needed.

(3) stack Gaussian noise

Watermark digit image shown in Fig. 3 a, Fig. 3 b, Fig. 3 c and Fig. 3 d is carried out noise jamming, it is that 0 variance is 0.05 Gaussian noise that noise is selected average, and watermark Bird, the Aimi that obtains, Flower and Scenery digital picture are shown in Figure 15 a, Figure 15 b, Figure 15 c and Figure 15 d.The bi-level digital watermark W ' of Figure 16 a for from the watermark Bird digital picture shown in Figure 15 a, extracting ₁, the bi-level digital watermark W ' of Figure 16 b for from the watermark Aimi digital picture shown in Figure 15 b, extracting ₂, the bi-level digital watermark W ' of Figure 16 c for from the watermark Flower digital picture shown in Figure 15 c, extracting ₃, the bi-level digital watermark W ' of Figure 16 d for from the watermark Scenery digital picture shown in Figure 15 d, extracting ₄Be that 0 variance is 0.05 Gauusian noise jammer for average; the visual quality generation serious degradation of watermark Bird, the Aimi shown in Figure 15 a, Figure 15 b, Figure 15 c and Figure 15 d, Flower and Scenery digital picture; Y-PSNR PSNR drops to respectively 20.0dB, 20.59dB, 20.07dB and 20.04dB; but by the watermark extracting result as can be known; the bi-level digital watermark that the present invention embeds has good anti-noise jamming ability; normalizated correlation coefficient r has reached respectively 0.969,0.839,0.977 and 0.967, can play the copyright protection effect.

With the original bi-level digital watermark W shown in Figure 17 a ₄As Query Information, watermark Bird, Aimi, Flower and Scenery digital picture shown in Figure 15 a, Figure 15 b, Figure 15 c and Figure 15 d to be retrieved, the result for retrieval that returns is shown in Figure 17 b, Figure 17 c, Figure 17 d and Figure 17 e.As seen, the watermark digit image although deteriroation of image quality is obvious, but still can retrieve user's results needed through after the Gauusian noise jammer quickly and accurately in the image library.

(4) how much cuttings

Watermark digit image shown in Fig. 3 a, Fig. 3 b, Fig. 3 c and Fig. 3 d is carried out the geometry cutting, cut 256 * 256 pixels in the upper right corner, obtain watermark Bird, Aimi, Flower and Scenery digital picture shown in Figure 18 a, Figure 18 b, Figure 18 c and Figure 18 d.The bi-level digital watermark W ' of Figure 19 a for from the watermark Bird digital picture shown in Figure 18 a, extracting ₁, the bi-level digital watermark W ' of Figure 19 b for from the watermark Aimi digital picture shown in Figure 18 b, extracting ₂, the bi-level digital watermark W ' of Figure 19 c for from the watermark Flower digital picture shown in Figure 18 c, extracting ₂, the bi-level digital watermark W ' of Figure 19 d for from the watermark Scenery digital picture shown in Figure 18 d, extracting ₄By the watermark extracting result as can be known; although watermark Bird, Aimi, Flower and Scenery digital picture are destroyed largely; Y-PSNR PSNR drops to 12.37dB, 10.63dB, 13.28dB and 11.68dB; but the inventive method has reasonable robustness for how much cuttings; the bi-level digital watermark that embeds still can be extracted well; normalizated correlation coefficient r has reached respectively 0.794,0.631,0.730 and 0.719, can play the copyright protection effect.

With the original bi-level digital watermark W shown in Figure 20 a ₁As Query Information, watermark Bird, Aimi, Flower and Scenery digital picture shown in Figure 18 a, Figure 18 b, Figure 18 c and Figure 18 d to be retrieved, the result for retrieval that returns is shown in Figure 20 b, Figure 20 c, Figure 20 d and Figure 20 e.As seen, although at this moment the watermark digit image geometric distortion is serious, but still can retrieve quickly and accurately user's results needed.

Claims (10)

1. data waterprint embedded method that is used for image retrieval and copyright protection is characterized in that its processing procedure is: at first, the original figure 8bit gray level image of bi-level digital watermark to be embedded is carried out normalized; Secondly, the digital picture that obtains after the normalized is carried out two-dimensional discrete wavelet conversion, obtain first a small echo ll channel and a plurality of the first wavelet details subgraph; The first small echo ll channel is decomposed into a fractional part the first small echo ll channel and integer part first a small echo ll channel; Again, pixel value according to each pixel in the bi-level digital watermark to be embedded, coefficient value scope in conjunction with each coefficient in fractional part the first small echo ll channel, coefficient value to each coefficient in fractional part the first small echo ll channel carries out the two-value quantification treatment, realize the embedding of bi-level digital watermark, and return a two-valued function table; Then, new fractional part the first small echo ll channel and integer part the first small echo ll channel that obtains after the two-value quantification treatment is combined into the first new small echo ll channel; The first new small echo ll channel and a plurality of the first wavelet details subgraph are carried out the 2-d discrete wavelet inverse transformation together, obtain embedding the digital picture of bi-level digital watermark; At last, the bi-level digital watermark is reached the logical table that in the bi-level digital watermark embed process, returns and be transferred to digital watermarking extraction end.

2. a kind of data waterprint embedded method for image retrieval and copyright protection according to claim 1 is characterized in that it specifically may further comprise the steps:

1.-1, at the digital watermarking built-in end, suppose to have K bi-level digital watermark to be embedded and the original figure 8bit gray level image of K bi-level digital watermark to be embedded, k bi-level digital watermark is designated as W _k, W _k={ w _k| w _k(m _k, n _k)=0 or 1,1≤m _k≤ M _k, 1≤n _k≤ N _k, the original figure 8bit gray level image of k bi-level digital watermark to be embedded is designated as F _k, F _k={ f _k| 0≤f _k(i _k, j _k)≤255,1≤i _k≤ I _k, 1≤j _k≤ J _k, wherein, 1≤k≤K, K 〉=1, W _kResolution be M _k* N _k, at this hypothesis M _k=N _k, w _k(m _k, n _k) expression W _kMiddle coordinate position is (m _k, n _k) the pixel value of pixel, F _kResolution be I _k* J _k, at this hypothesis I _k=J _k, f _k(i _k, j _k) expression F _kMiddle coordinate position is (i _k, j _k) the pixel value of pixel;

1.-2 the original figure 8bit gray level image of, supposing the current bi-level digital watermark to be embedded of processing is the original figure 8bit gray level image F of k bi-level digital watermark to be embedded _k

1.-3, to F _kCarry out normalized, with F _kThe digital picture F ' that after normalized, obtains _kMiddle coordinate position is (i _k, j _k) the pixel value of pixel be designated as f ' _k(i _k, j _k), f ' _k(i _k, j _k)=f _k(i _k, j _k)/255, wherein, 1≤i _k≤ I _k, 1≤j _k≤ J _k

1.-4, to F ' _kCarry out L _kThe level two-dimensional discrete wavelet conversion obtains first a small echo ll channel and a plurality of the first wavelet details subgraph, with F ' _kThe first corresponding small echo ll channel is designated as FA _k, wherein, L _k=log ₂(I _k/ M _k)=log ₂(J _k/ N _k), FA _kResolution be

1.-5, with FA _kBe decomposed into fractional part and integer part two parts data subgraph, obtain a fractional part the first small echo ll channel and integer part first a small echo ll channel, be designated as respectively FAD _kAnd FAI _k

1.-6, according to W _kIn the pixel value of each pixel, to FAD _kIn the coefficient value of each coefficient carry out the two-value quantification treatment, realize bi-level digital watermark W _kEmbedding, and to return a size be M _k* N _kThe two-valued function table, the fractional part first small echo ll channel that will obtain after the two-value quantification treatment is designated as FAD ' _k, be Lg with the two-valued function souvenir that returns _k

1.-7, with FAD ' _kAnd FAI _kBe combined into F ' _kCorresponding the first new small echo ll channel is designated as FA ' _k

1.-8, with FA ' _kAnd F ' _kCorresponding a plurality of the first wavelet details subgraphs carry out L together again _kThe inverse transformation of level 2-d discrete wavelet, obtaining embedding has bi-level digital watermark W _kDigital picture, be designated as F " _k

1.-9, make k=k+1, then returning step 1.-2 continues the original figure 8bit gray level image of next bi-level digital watermark to be embedded is processed, until the original figure 8bit gray level image of K bi-level digital watermark to be embedded all is disposed, wherein, "=" among the k=k+1 is assignment;

1.-10, with K bi-level digital watermark W ₁, W ₂..., W _k..., W _KAnd the two-valued function table Lg that in the bi-level digital watermark embed process, returns ₁, Lg ₂..., Lg _k..., Lg _KBe transferred to digital watermarking and extract end, wherein, W ₁Represent the 1st bi-level digital watermark, W ₂Represent the 2nd bi-level digital watermark, W _KRepresent K bi-level digital watermark, Lg ₁Be illustrated in bi-level digital watermark W ₁The two-valued function table that returns in the telescopiny, Lg ₂Be illustrated in bi-level digital watermark W ₂The two-valued function table that returns in the telescopiny, Lg _KBe illustrated in bi-level digital watermark W _KThe two-valued function table that returns in the telescopiny.

3. a kind of data waterprint embedded method for image retrieval and copyright protection according to claim 2 is characterized in that making during described step 1.-5

Expression FA _kMiddle coordinate position is

The coefficient value of coefficient, order Expression FAD _kMiddle coordinate position is

The coefficient value of coefficient, order

Expression FAI _kMiddle coordinate position is

The coefficient value of coefficient, then have

Wherein, symbol

The maximum integer less than himself is got in expression, $1\≤\widetilde{i_k}\≤(I_k/2^{L_k}),$ $1\≤\widetilde{j_k}\≤(J_k/2^{L_k}).$

4. according to claim 2 or 3 described a kind of data waterprint embedded methods for image retrieval and copyright protection, it is characterized in that described step detailed process 1.-6 is: a1, with FAD _kIn the current Coefficient Definition of processing be current coefficient; A2, suppose that the coordinate position of current coefficient is (m _k, n _k), then the coefficient value with current coefficient is designated as fad _k(m _k, n _k), wherein, 1≤m _k≤ M _k, 1≤n _k≤ N _kA3, according to fad _k(m _k, n _k) and W _kMiddle coordinate position is (m _k, n _k) the pixel value w of pixel _k(m _k, n _k), calculate FAD ' _kMiddle coordinate position is (m _k, n _k) the coefficient value of coefficient, be designated as fad ' _k(m _k, n _k), and return two-valued function table Lg _kMiddle coordinate position is (m _k, n _k) the logical value lg of element _k(m _k, n _k), specific as follows: as to work as w _k(m _k, n _k)=0 and fad _k(m _k, n _k) ∈ [0,0.5) time, with fad _k(m _k, n _k) be quantified as 0.25, i.e. fad ' _k(m _k, n _k)=0.25, lg _k(m _k, n _k)=0; Work as w _k(m _k, n _k)=0 and fad _k(m _k, n _k) ∈ [0.5,1), with fad _k(m _k, n _k) be quantified as 0.75, i.e. fad ' _k(m _k, n _k)=0.75, lg _k(m _k, n _k)=1; Work as w _k(m _k, n _k)=1 and fad _k(m _k, n _k) ∈ [0,0.5), with fad _k(m _k, n _k) be quantified as 0.25, i.e. fad ' _k(m _k, n _k)=0.25, lg _k(m _k, n _kW is worked as in)=1 _k(m _k, n _k)=1 and fad _k(m _k, n _k) ∈ [0.5,1), with fad _k(m _k, n _k) be quantified as 0.75, i.e. fad ' _k(m _k, n _k)=0.75, lg _k(m _k, n _k)=0; A4, with FAD _kThen the pending coefficient of the middle next one returns step a2 and continues to carry out, until FAD as current coefficient _kIn all coefficient processing complete, realize bi-level digital watermark W _kEmbedding, obtain fractional part the first small echo ll channel FAD ' that after the two-value quantification treatment, obtains _k

5. a kind of data waterprint embedded method for image retrieval and copyright protection according to claim 4 is characterized in that described step detailed process 1.-7 is: b1, with FAD ' _kIn the current Coefficient Definition of processing be current the first coefficient, with FAI _kIn the current Coefficient Definition of processing be current the second coefficient; B2, suppose that the coordinate position of current the first coefficient is Then the coefficient value with current the first coefficient is designated as

The coordinate position of supposing current the second coefficient is

Then the coefficient value with current the second coefficient is designated as Wherein,

B3, basis

With

Calculate FA ' _kMiddle coordinate position is

The coefficient value of coefficient, be designated as

B4, with FAD ' _kThe pending coefficient of the middle next one is as current the first coefficient, with FAI _kThen the pending coefficient of the middle next one returns step b2 and continues to carry out, until FAD ' as current the second coefficient _kAnd FAI _kIn all coefficient processing complete, obtain F ' _kCorresponding the first new small echo ll channel FA ' _k

6. digital watermarking extracting method that is used for image retrieval and copyright protection is characterized in that its processing procedure is: at first, the digital 8bit gray level image of bi-level digital watermark to be extracted is carried out normalized; Secondly, the digital picture that obtains after the normalized is carried out two-dimensional discrete wavelet conversion, obtain second a small echo ll channel and a plurality of the second wavelet details subgraph; The second small echo ll channel is decomposed into a fractional part the second small echo ll channel and integer part second a small echo ll channel; Again, be transferred to the two-valued function table that end is extracted in digital watermarking according to the digital watermarking built-in end, and the coefficient value scope of each coefficient in fractional part the second small echo ll channel, carry out two-value and differentiate, extract the bi-level digital watermark; Then, be transferred to the coefficient correlation that bi-level digital watermark with the bi-level digital watermark that extracts of end are extracted in digital watermarking by calculating the digital watermarking built-in end, determine whether the bi-level digital watermark that extracts is the bi-level digital watermark that is used for copyright protection that the digital watermarking built-in end embeds; At last, utilize coefficient correlation to realize the retrieval of digital picture.

7. a kind of digital watermarking extracting method for image retrieval and copyright protection according to claim 6 is characterized in that it specifically may further comprise the steps:

2.-1, extract end in digital watermarking, suppose to have the individual bi-level digital watermark to be extracted of K and the digital 8bit gray level image of K bi-level digital watermark to be extracted, k bi-level digital watermark is designated as W ' _k, W ' _k=w ' _k| w ' _k(m ' _k, n ' _k)=0 or 1,1≤m ' _k≤ M ' _k, 1≤n ' _k≤ N ' _k, the digital 8bit gray level image of k bi-level digital watermark to be extracted is designated as TF _k, TF _k={ tf _k| 0≤tf _k(i ' _k, j ' _k)≤255,1≤i ' _k≤ I ' _k, 1≤j ' _k≤ J ' _k, wherein, 1≤k≤K, K 〉=1, W ' _kResolution be M ' _k* N ' _k, at this hypothesis M ' _k=N ' _k, and W ' _kThe resolution of the bi-level digital watermark that embeds of resolution and digital watermarking built-in end identical, w ' _k(m ' _k, n ' _k) expression W ' _kMiddle coordinate position be (m ' _k, n ' _k) the pixel value of pixel, TF _kResolution be I ' _k* J ' _k, at this hypothesis I ' _k=J ' _k, and TF _kResolution identical with the resolution of the original figure 8bit gray level image of digital watermarking built-in end bi-level digital watermark to be embedded, tf _k(i ' _k, j ' _k) expression TF _kMiddle coordinate position be (i ' _k, j ' _k) the pixel value of pixel;

2.-2 the digital 8bit gray level image of, supposing the current bi-level digital watermark to be extracted of processing is the digital 8bit gray level image TF of k bi-level digital watermark to be extracted _k

2.-3, to TF _kCarry out normalized, with TF _kThe digital picture TF ' that after normalized, obtains _kMiddle coordinate position be (i ' _k, j ' _k) the pixel value of pixel be designated as tf ' _k(i ' _k, j ' _k), tf ' _k(i ' _k, j ' _k)=tf _k(i ' _k, j ' _k)/255, wherein, 1≤i ' _k≤ I ' _k, 1≤j ' _k≤ J ' _k

2.-4, to TF ' _kCarry out L ' _kThe level two-dimensional discrete wavelet conversion obtains second a small echo ll channel and a plurality of the second wavelet details subgraph, with TF ' _kThe second corresponding small echo ll channel is designated as TFA _k, wherein, L ' _k=log ₂(I ' _k/ M ' _k)=log ₂(J ' _k/ N ' _k), TFA _kResolution be

2.-5, with TFA _kBe decomposed into fractional part and integer part two parts data subgraph, obtain a fractional part the second small echo ll channel and integer part second a small echo ll channel, be designated as respectively TFAD _kAnd TFAI _k

2.-6, be transferred to the two-valued function table Lg that end is extracted in digital watermarking according to the digital watermarking built-in end _kAnd TFAD _kIn the coefficient value of each coefficient, extract bi-level digital watermark W ' _k

2.-7, calculate the digital watermarking built-in end and be transferred to the bi-level digital watermark W that end is extracted in digital watermarking _kWith the bi-level digital watermark W ' that extracts _kNormalizated correlation coefficient, be designated as r, r=corr (W _k, W ' _k), then determine according to the size of r whether the bi-level digital watermark Wk that is used for copyright protection that the digital watermarking built-in end embeds is successfully extracted, wherein, the computing function of corr () expression normalizated correlation coefficient;

2.-8, make k=k+1, then returning step 2.-2 continues the digital 8bit gray level image of next bi-level digital watermark to be extracted is processed, until the digital 8bit gray level image of K bi-level digital watermark to be extracted all is disposed, wherein, "=" among the k=k+1 is assignment;

2.-9, be transferred to K the bi-level digital watermark W that end is extracted in digital watermarking from the digital watermarking built-in end ₁, W ₂..., W _k..., W _KIn select arbitrarily a bi-level digital watermark as Query Information, should optional bi-level digital watermark be designated as W _z, then calculate respectively W _zWith K bi-level digital watermark W ' that extracts ₁, W ' ₂..., W ' _k..., W ' _KNormalizated correlation coefficient, correspondence is designated as r ₁, r ₂..., r _k..., r _K, r ₁=corr (W _z, W ' ₁), r ₂=corr (W _z, W ' ₂), r _k=corr (W _z, W ' _k), r _K=corr (W _z, W ' _K), at last according to K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KReturn the digital 8bit gray level image that the corresponding K width of cloth embeds the bi-level digital watermark by large and little order, realize the retrieval of digital picture, wherein, 1≤z≤K, W ' ₁The 1st the bi-level digital watermark that expression extracts, W ' ₂The 2nd the bi-level digital watermark that expression extracts, W ' _kK the bi-level digital watermark that expression extracts, W ' _KK the bi-level digital watermark that expression extracts.

8. a kind of digital watermarking extracting method for image retrieval and copyright protection according to claim 7 is characterized in that making during described step 2.-5

Expression TFA _kMiddle coordinate position is

The coefficient value of coefficient, order

Expression TFAD _kMiddle coordinate position is

The coefficient value of coefficient, order

Expression TFAI _kMiddle coordinate position is

The coefficient value of coefficient, then have

Wherein, symbol The maximum integer less than himself is got in expression, $1\≤\widetilde{i_k^{\′}}\≤(I_k^{\′}/2^{{L_k}^{\′}}),$ $1\≤\widetilde{j_k^{\′}}\≤(J_k^{\′}/2^{{L_k}^{\′}});$

Described step detailed process 2.-6 is: c1, with TFAD _kIn the current Coefficient Definition of processing be current coefficient, with Lg _kIn the current element definition of processing be currentElement; C2, suppose current coefficient coordinate position for (m ' _k, n ' _k), then the coefficient value with current coefficient is designated as tfad _k(m ' _k, n ' _k), the coordinate position of supposing currentElement for (m ' _k, n ' _k), then the logical value with currentElement is designated as lg _k(m ' _k, n ' _k), wherein, 1≤m ' _k≤ M ' _k, 1≤n ' _k≤ N ' _kC3, according to tfad _k(m ' _k, n ' _k) and lg _k(m ' _k, n ' _k), extract bi-level digital watermark W ' _kMiddle coordinate position be (m ' _k, n ' _k) the pixel value of pixel, be designated as w ' _k(m ' _k, n ' _k), specific as follows: as to work as lg _k(m ' _k, n ' _k)=0 and tfad _k(m ' _k, n ' _k) ∈ [0,0.5) time, w ' put _k(m ' _k, n ' _k)=0; Work as lg _k(m ' _k, n ' _k)=0 and tfad _k(m ' _k, n ' _k) ∈ [0.5,1) time, w ' put _k(m ' _k, n ' _k)=1; Work as lg _k(m ' _k, n ' _k)=1 and tfad _k(m ' _k, n ' _k) ∈ [0,0.5) time, w ' put _k(m ' _k, n ' _k)=1; Work as lg _k(m ' _k, n ' _k)=1 and tfad _k(m ' _k, n ' _k) ∈ [0.5,1) time, w ' put _k(m ' _k, n ' _k)=0; C4, with TFAD _kThe pending coefficient of the middle next one is as current coefficient, with Lg _kThen the pending element of the middle next one returns step c2 and continues to carry out, until TFAD as currentElement _kIn all coefficients and Lg _kIn all elements be disposed, realize bi-level digital watermark W ' _kExtraction.

9. according to claim 7 or 8 described a kind of digital watermarking extracting methods for image retrieval and copyright protection, it is characterized in that determining the bi-level digital watermark W that is used for copyright protection that the digital watermarking built-in end embeds according to the size of r during described step 2.-7 _kThe detailed process of whether successfully being extracted is: if the value of r is 1, then determine the bi-level digital watermark W ' that extracts _kBi-level digital watermark W with the embedding of digital watermarking built-in end _kIn full accord, show the bi-level digital watermark W that the digital watermarking built-in end embeds _kNondestructively extracted, be can be used for copyright protection; If the value of r is between 1～δ _TBetween, then determine the bi-level digital watermark W ' that extracts _kBi-level digital watermark W with the embedding of digital watermarking built-in end _kSimilitude very large, this moment can be from the bi-level digital watermark W ' that extracts _kIn recognize the bi-level digital watermark W that the digital watermarking built-in end embeds _k, show the bi-level digital watermark W that the digital watermarking built-in end embeds _kSuccessfully extracted, be can be used for copyright protection, wherein, δ _TBe the watermark extracting thresholding of setting; If the value of r is less than δ _T, then determine the bi-level digital watermark W ' that extracts _kBi-level digital watermark W with the embedding of digital watermarking built-in end _kSimilitude very little, this moment can't be from the bi-level digital watermark W ' that extracts _kIn recognize the bi-level digital watermark W that the digital watermarking built-in end embeds _k, show the bi-level digital watermark W that the digital watermarking built-in end embeds _kExtract unsuccessfully.

10. a kind of digital watermarking extracting method for image retrieval and copyright protection according to claim 9 is characterized in that during described step 2.-9 according to K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KReturning the detailed process that the corresponding K width of cloth embeds the digital 8bit gray level image that the bi-level digital watermark is arranged by large and little order is: d1, find out K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KA normalizated correlation coefficient of intermediate value maximum is designated as maxr ₁, suppose maxr ₁Be K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KIn the individual normalizated correlation coefficient of k ', then returning to embed has the individual bi-level digital watermark of k ' W _{K '}Digital 8bit gray level image, as the 1st width of cloth image that meets the user search requirement, wherein, 1≤k '≤K; D2, find out a normalizated correlation coefficient of remaining K-1 normalizated correlation coefficient intermediate value maximum, be designated as maxr ₂, suppose maxr ₂Be K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KIn k " individual normalizated correlation coefficient, then returning to embed has k " individual bi-level digital watermark W _{K "}Digital 8bit gray level image, as the 2nd width of cloth image that meets the user search requirement, wherein, 1≤k "≤K; D3, find out a normalizated correlation coefficient of remaining K-2 normalizated correlation coefficient intermediate value maximum, be designated as maxr ₃, suppose maxr ₃Be K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KIn k ' " individual normalizated correlation coefficient, then returning to embed has k ' " individual bi-level digital watermark W _{K ' "}Digital 8bit gray level image, as the 3rd width of cloth image that meets the user search requirement, wherein, 1≤k ' "≤K; D4, according to the operation of steps d 3, continue in an identical manner to find out a normalizated correlation coefficient of each remaining normalizated correlation coefficient intermediate value maximum, return the digital 8bit gray level image that corresponding embedding has the bi-level digital watermark with this, until remaining last normalizated correlation coefficient is designated as maxr with last normalizated correlation coefficient _K, suppose maxr _KBe K normalizated correlation coefficient r ₁, r ₂..., r _k..., r _KIn k* normalizated correlation coefficient, then returning to embed has k* bi-level digital watermark W _K*Digital 8bit gray level image, the last piece image that arrives as user search, wherein, 1≤k*≤K.

Images