CN102682418A - Method for embedding and extracting multiple zero watermarks of digital image - Google Patents

Abstract

The invention discloses a method for embedding and extracting multiple zero watermarks of a digital image. The embedding and extraction of multiple zero watermarks are all carried out in the compound domain of discrete wavelet transform and discrete cosine transform. The embedding of multiple zero watermarks utilizes the original digital image The binary watermark key is constructed by the coefficients in the composite domain, and then XOR operation is carried out with the actual multiple binary digital watermarks. The extraction of multiple zero watermarks is to use the coefficients in the composite domain of the digital image to be tested to construct the binary watermark key. Combined with zero watermark information, the embedding and extraction method of the present invention has excellent watermark robustness and can resist conventional image processing attacks; at the same time, the embedding method of the present invention registers multiple binary digital watermarks into the copyright database, which is not correct No changes are made to the original digital image, so the embedded multiple binary digital watermarks are completely imperceptible, which not only balances the contradiction between the robustness and imperceptibility of digital watermarking, but also satisfies the diverse needs of digital watermarking applications.

Description

A Multiple Zero-Watermark Embedding and Extraction Method for Digital Image

technical field

The invention relates to a zero watermark technology, in particular to a method for embedding and extracting multiple zero watermarks of a digital image.

Background technique

As a technical means to effectively solve the copyright protection and ownership identification of digital media works, digital watermarking is a research hotspot in the field of digital media information security. In order to play its due role, digital watermarking must have two basic elements: robustness and imperceptibility. Watermark robustness means that after the digital media undergoes conventional digital signal processing or external attacks, the embedded digital watermark still has good detectability or can still reflect the copyright information of the original digital media. The imperceptibility of watermark means that the embedding of digital watermark cannot affect the auditory or visual quality of the original digital media, so it will not affect the application value of the original digital media. Obviously, the traditional digital watermarking technology, by modifying the spatial domain data or transform domain coefficients of the original digital media, will include some characteristic information of the author or work, such as signature, copyright logo, serial number, date or icon, etc. When digital watermarking is embedded in the original digital media, there is inevitably a contradiction between watermark robustness and watermark imperceptibility: on the one hand, watermark robustness requires embedding as much digital watermark information as possible in the original digital media to Resist various conventional processing or intentional attacks, and on the other hand, watermark imperceptibility hopes to embed as little digital watermark information as possible to avoid obvious differences from the original digital media, which is important for some sensitive digital images such as medical images and Palmprint images are particularly important, because the detailed pixels of such digital images contain extremely important information, and the distortion caused by any pixel change in such digital images will affect the judgment of the original digital image.

In recent years, the zero watermark technology has been proposed to solve the contradiction between watermark robustness and watermark imperceptibility in traditional digital watermark technology, and has become a new research branch in digital watermark technology. The so-called zero watermark technology is to construct a watermark key through the characteristics of the original digital media, and then combine it with the digital watermark to be embedded to form relevant zero watermark information, and then register it in the intellectual property watermark information database. Watermarking technology, which does not make any modification to the original digital media. Therefore, in the zero watermarking technology, the construction and registration process of the zero watermark is the digital watermark embedding process in the usual sense. Once the registration is completed, the original digital media is considered to contain the actual digital watermark and has corresponding copyright protection capabilities. Since the actual digital watermark in the zero watermarking technology is registered in the intellectual property watermark information database instead of being embedded in the original digital media, there is no problem of degradation in the quality of the original digital media. Because of this feature, the zero watermarking technology is very It well balances the contradiction between watermark robustness and watermark imperceptibility in traditional digital watermarking technology.

Currently, several zero-watermarking schemes have been proposed. In 2008, Zeng Fanjuan and Zhou Anmin proposed a digital image zero-watermarking scheme based on Contourlet transform and singular value decomposition in the Computer Application Journal. The scheme first performs Contourlet transform on the original image and decomposes it into a series of multi-scale, local Then select the low-frequency sub-band for block singular value decomposition, and construct the zero watermark according to the invariance of the integer bit size of the first singular value in each block decomposition. In 2009, Zhao Jie, Wang Hao and He Bing proposed a zero-watermarking scheme based on image scrambling and wavelet transform in the Journal of Computer Engineering and Science. The low-frequency sub-image after first-level wavelet decomposition of the image and the watermark image after scrambling are extracted matrix, and then the extracted matrix is scrambled and encrypted to complete the embedding and extraction of zero watermark. In 2010, Bi Xiuli, He Chunxiang and Cheng Cheng proposed a zero-watermarking scheme based on logarithmic polar coordinate mapping and wavelet lifting in the Journal of Computer Engineering and Science. Scaling and normalization processing, then logarithmic polar coordinate mapping, and then one-level integer wavelet lifting, and finally using the decomposed low-frequency sub-image and the scrambled watermark image to construct a zero watermark. In 2011, Wang Wujun proposed a zero-watermarking scheme of multi-level discrete cosine transform and singular value decomposition in the Journal of Computer and Digital Engineering. The scheme first performs multi-level discrete cosine transform on the original image to be embedded, extracts low-frequency sub-images and Singular value decomposition is performed on it, and then the embedded watermark image is scrambled, and the scrambled watermark image is subjected to singular value decomposition. Finally, the original image and the watermark image are calculated by the two singular value decomposition results to generate a zero watermark.

However, the zero watermark schemes that have been proposed so far are basically based on the digital watermark technology of single zero watermark embedding and extraction, and have the limitation of single function. In practical applications, there are more needs to embed multiple digital watermarks into digital media works to meet different application purposes. For example, after a digital media work is completed, it needs two or more A digital media work needs to mark the legal information used by different owners (such as creators, distributors, users, etc.) at different stages such as release, sale, and use. Simultaneously embed a robust watermark for copyright authentication and a fragile watermark for content authentication, etc. Therefore, considering the application diversity of digital watermarking, the research of zero-watermarking scheme embedded with multiple digital watermarks has more practical application value and broad application prospects.

Contents of the invention

The technical problem to be solved by the present invention is to provide a method for embedding and extracting multiple zero watermarks of a digital image, which realizes embedding multiple zero watermarks in the original digital image, and the embedded multiple zero watermarks have excellent robust performance , while effectively ensuring that the quality of the original digital image is not affected.

The technical solution adopted by the present invention to solve the above-mentioned technical problems is: a method for embedding multiple zero-watermarks of a digital image, which is characterized in that it comprises the following steps:

①-1. At multiple zero-watermark embedding terminals, assume that the original digital image to be embedded with K binary digital watermarks is an 8-bit grayscale image, and denoted as F, F={0≤f(m,n)≤255 ,1≤m≤M,1≤n≤N}, where K≥2, M represents the vertical resolution of the original digital image F to be embedded with K binary digital watermarks, and N represents the K binary digital watermarks to be embedded The horizontal resolution of the original digital image F of the watermark, M×N represents the resolution of the original digital image F to be embedded with K binary digital watermarks, f(m,n) represents the original digital image F to be embedded with K binary digital watermarks The pixel value of the pixel whose coordinate position is (m, n) in the image F;

①-2. At multiple zero-watermark embedding terminals, assuming that the K binary digital watermarks to be embedded are all binary images, which are recorded as W ₁ , W ₂ ,...,W _k ,...W _K , then for the The embedded kth binary digital watermark W _k , W _k ={w _k (i _k , j _k )=0 or 1, 1≤i _k ≤I _k ,1≤j _k ≤J _k }, where, 1 ≤k≤K, I _k represents the vertical resolution of the kth binary digital watermark W _k to be embedded, J _k represents the horizontal resolution of the kth binary digital watermark W _k to be embedded, I _k ×J _k represents the resolution of the kth binary digital watermark W _k to be embedded, w _k (i _k , j _k ) represents the coordinate position of the kth binary digital watermark W _k to be embedded is (i _k , j _k ) pixel value of the pixel;

①-3. The original digital image F to be embedded with K binary digital watermarks is normalized to obtain the normalized digital image, denoted as F′, and the normalized digital image F′ is The pixel value of the pixel whose coordinate position is (m,n) is recorded as f'(m,n), f'(m,n)=f(m,n)/255;

min()为取最小值函数，max()为取最大值函数，符号表示取小于其自身的最大整数，I₁表示待嵌入的第1个二值数字水印W₁的竖直分辨率，J₁表示待嵌入的第1个二值数字水印W₁的横向分辨率，I_K表示待嵌入的第K个二值数字水印W_K的竖直分辨率，J_K表示待嵌入的第K个二值数字水印W_K的横向分辨率；①-4. Carry out L-level two-dimensional discrete wavelet transform to F′ to obtain a first wavelet approximation subgraph and multiple first wavelet detail subgraphs, and denote the first wavelet approximation subgraph as FA, where the resolution of FA The rate is (M/2 ^L )×(N/2 ^L ),

min() is the minimum value function, max() is the maximum value function, the symbol Indicates the largest integer smaller than itself, I ₁ represents the vertical resolution of the first binary digital watermark W ₁ to be embedded, J ₁ represents the horizontal resolution of the first binary digital watermark W ₁ to be embedded, I _K represents the vertical resolution of the Kth binary digital watermark W _K to be embedded, and J _K represents the horizontal resolution of the Kth binary digital watermark W _K to be embedded;

①-5. Carry out two-dimensional discrete cosine transform to FA to obtain a first two-dimensional discrete cosine transform coefficient matrix with the same resolution as FA, denoted as FAC, and then carry out Zig-Zag scanning arrangement on FAC to obtain a first one dimensional discrete cosine transform coefficient sequence, denoted as FACS, FACS={facs(x),1≤x≤(M/2 ^L )×(N/2 ^L )}, where facs(x) represents the x-th in FACS Discrete cosine transform coefficients, the second discrete cosine transform coefficients in FACS are initially all discrete cosine transform AC coefficients;

①-6. According to the resolution of each binary digital watermark to be embedded, select I ₁ ×J ₁ , I ₂ ×J ₂ ,…, I _k ×J _k , ...and I _K ×J _K discrete cosine transform AC coefficients to form K first one-dimensional discrete cosine transform AC coefficient sequences, respectively denoted as FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K , for from The kth first one-dimensional discrete cosine transform AC coefficient sequence FACS k composed of I _k ×J _k discrete cosine transform AC coefficients selected in FACS that _{meet the set conditions, FACS k =} {facs _k (y),1 ≤y≤I _k ×J _k }, and then record the corresponding position information of each discrete cosine transform AC coefficient in FACS in FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K , where I ₂ represents the The vertical resolution of the second binary digital watermark W ₂ , J ₂ represents the horizontal resolution of the second binary digital watermark W ₂ to be embedded, and facs _k (y) represents the yth discrete value in FACS _k For the cosine transform AC coefficient, the setting condition is such that the absolute value of the difference between any two adjacent discrete cosine transform AC coefficients in the formed first one-dimensional discrete cosine transform AC coefficient sequence is greater than or equal to the set difference threshold;

①-7. Compare the size of any two adjacent discrete cosine transform AC coefficients in FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K respectively, and return a logical value 1 or 0 according to the comparison result, for FACS _k Any two adjacent discrete cosine transform AC coefficients in , record them as facs _k (z) and facs _k (z+1) respectively, and judge whether facs _k (z)>facs _k (z+1) is true, if If it is true, return a logic value of 1, otherwise, return a logic value of 0, where, 1≤z≤I _k ×J _k -1; then compare the last one of FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K DCT AC coefficient and the size of the first DCT AC coefficient, and return a logic value of 1 or 0 according to the comparison result. For the last DCT AC coefficient and the first DCT AC coefficient in FACS _k , If _the former is greater, return _a logical value of 1, otherwise, return a logical value of ₀ ; then construct a one-to-one correspondence of the first and _second value digital watermark key, for the returned logical value corresponding to FACS _k , the returned logical value is stored in a two-dimensional matrix with a size of I _k × J _k in the order of first row and second column, and the two-dimensional matrix is used as The k-th first binary digital watermark key, denoted as WB _k ;

①-8. The K binary digital watermarks W ₁ , W ₂ ,...,W _k ,...,W _K to be embedded are respectively scrambled, and the K binary digital watermarks obtained after the scrambled processing are respectively corresponding to are WS ₁ , WS ₂ , ..., WS _k , ... and _WSK , then combine WS ₁ , WS ₂ , ..., WS _k , ... and WS _K with K first binary digital watermark keys WB ₁ , WB ₂ , ..., WB _k , ... and WB _K are one-to-one XORed to obtain K zero-watermark information, respectively recorded as WO ₁ , WO ₂ , ..., WO _k , ... and WO _K , WO ₁ =xor( WS ₁ ,WB ₁ ), WO ₂ =xor(WS ₂ ,WB ₂ ),…, WO _k =xor(WS _k ,WB _k ),…, WO _K =xor(WS _K ,WB _K ), and K Zero watermark information WO ₁ , WO ₂ , ..., WO _k , ... and WO _K are stored in the digital watermark information database of the registration authority, and K binary digital watermarks W ₁ , W ₂ , ..., W _k , ..., The embedding of W _K , where WS ₁ represents the binary digital watermark obtained after scrambling W ₁ , WS ₂ represents the binary digital watermark obtained after scrambling W ₂ , and WS _k represents the binary watermark obtained after scrambling W _k The binary digital watermark obtained after scrambling processing, WS _K represents the binary digital watermark obtained after scrambling W _K , WB ₁ represents the first first binary digital watermark key, WB ₂ represents the second The first binary digital watermark key, WB _K represents the Kth first binary digital watermark key, and xor() is an exclusive OR operation function;

①-9. At multiple zero-watermark embedding terminals, record the position information of each discrete cosine transform AC coefficient in FACS in FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K , and K zero-watermark information WO ₁ , WO ₂ ,..., WO _k ,..., WO _K and K binary digital watermarks W ₁ , W ₂ ,..., W _k ,..., W _K are transmitted to multiple zero-watermark extraction terminals.

The absolute value of the difference between any two adjacent discrete cosine transform AC coefficients in FACS ₁ , FACS ₂ , ... , FACS _k , ... and FACS _K in steps ①-6 is greater than or equal to δ ₁ , δ ₂ respectively , ..., δ _k , ... and δ _K , where δ ₁ represents the first difference threshold set for the absolute value of the difference between any two adjacent discrete cosine transform AC coefficients in FACS ₁ , and δ ₂ represents The second difference threshold set for the absolute value of the difference between any two adjacent discrete cosine transform AC coefficients in FACS ₂ , δ _k represents the value of any two adjacent discrete cosine transform AC coefficients in FACS _k The kth difference threshold set by the absolute value of the difference, δ _K represents the Kth difference threshold set for the absolute value of the difference between any two adjacent discrete cosine transform AC coefficients in FACS _K.

A method for extracting multiple zero watermarks of a digital image, characterized in that it comprises the following steps:

②-1. At multiple zero-watermark extraction terminals, record the digital images of K binary digital watermarks to be extracted as TF, TF={0≤tf(m′,n′)≤255, 1≤m′≤M ', 1≤n'≤N'}, where K≥2, M' represents the vertical resolution of the digital image TF of the K binary digital watermarks to be extracted, and N' represents the resolution of the K binary digital watermarks to be extracted The horizontal resolution of the digital image TF, M'×N' represents the resolution of the digital image TF to be extracted K binary digital watermarks, the resolution of the digital image TF to be extracted K binary digital watermarks and multiple zero watermarks The resolution of the digital image embedded with the digital watermark at the embedding end is the same, and tf(m',n') represents the pixel whose coordinate position is (m',n') in the digital image TF of the K binary digital watermark to be extracted value;

②-2. At multiple zero-watermark extraction terminals, mark the K binary digital watermarks to be extracted as W′ ₁ , W′ ₂ , ..., W′ _k , ... and W′ _K . k binary digital watermarks W′ _k , W′ _k ={w′ _k (i′ _k ,j′ _k )=0 or 1, 1≤i′ _k ≤I′ _k ,1≤j′ _k ≤J′ _k }, where, 1≤k≤K, I′ _k represents the vertical resolution of the kth binary digital watermark W′ _k to be extracted, and J′ _k represents the kth binary digital watermark W′ to be extracted The horizontal resolution of _k , I′ _k ×J′ _k represents the resolution of the kth binary digital watermark W′ _k to be extracted, and the resolution of the kth binary digital watermark W′ _k to be extracted is related to multiple The resolution of the kth binary digital watermark embedded in the zero watermark embedding end is the same, w′ _k (i′ _k , j′ _k ) means that the coordinate position of the kth binary digital watermark W′ _k to be extracted is (i ′ _k , j′ _k ) pixel value of the pixel;

②-3. Perform normalization processing on the digital image TF to be extracted K binary digital watermarks, and obtain the digital image after normalization processing, which is denoted as TF′, and the coordinates in the digital image TF′ after normalization processing The pixel value of the pixel at position (m',n') is recorded as tf'(m',n'), tf'(m',n')=tf(m',n')/255;

min()为取最小值函数，max()为取最大值函数，符号

表示取小于其自身的最大整数，I₁′表示待提取的第1个二值数字水印W₁′的竖直分辨率，J₁′表示待提取的第1个二值数字水印W₁′的横向分辨率，I_K′表示待提取的第K个二值数字水印W′_K的竖直分辨率，J_K′表示待提取的第K个二值数字水印W′_K的横向分辨率；②-4. Carry out L'-level two-dimensional discrete wavelet transform to TF' to obtain a second wavelet approximation subgraph and a plurality of second wavelet detail subgraphs, and denote the second wavelet approximation subgraph as TFA, wherein the TFA's The resolution is (M′/2 ^L ′)×(N′/2 ^L ′),

min() is the minimum value function, max() is the maximum value function, the symbol

Indicates the largest integer smaller than itself, I ₁ ′ represents the vertical resolution of the first binary digital watermark W ₁ ′ to be extracted, J ₁ ′ represents the resolution of the first binary digital watermark W ₁ ′ to be extracted Horizontal resolution, I _K ′ represents the vertical resolution of the Kth binary digital watermark W′ _K to be extracted, and J _K ′ represents the horizontal resolution of the Kth binary digital watermark W′ _K to be extracted;

②-5. Perform two-dimensional discrete cosine transform on TFA to obtain a second two-dimensional discrete cosine transform coefficient matrix with the same resolution as TFA, denoted as TFAC, and then carry out Zig-Zag scanning arrangement on TFAC to obtain a second one dimensional discrete cosine transform coefficient sequence, denoted as TFACS, TFACS={tfacs(x′),1≤x′≤(M′/2 ^L ′)×(N′/2 ^L ′)}, where, tfacs(x′ ) represents the x′th discrete cosine transform coefficient in TFACS, and the second discrete cosine transform coefficient in TFACS is initially an AC discrete cosine transform coefficient;

②-6. According to the corresponding position information in FACS of each discrete cosine transform AC coefficient in FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K recorded by multiple zero-watermark embedding terminals, extract from TFACS respectively The I ₁ ′×J ₁ ′, I ₂ ′×J ₂ ′, ..., I _k ′×J _k ′, ... and I _K ′×J _K ′ discrete cosine transform AC coefficients at the corresponding positions constitute K second The one-dimensional discrete cosine transform AC coefficient sequence is correspondingly denoted as TFACS ₁ , TFACS ₂ , ..., TFACS _k , ... and TFACS _K , for each discrete cosine transform AC coefficient in FACS _k recorded according to multiple zero-watermark embedding terminals in Corresponding position information in FACS, extract the kth second one-dimensional discrete cosine transform AC coefficient sequence TFACS k composed of I _k ′×J _k ′ discrete cosine transform AC coefficients corresponding to the position from TFACS, TFACS _{k =} { tfacs _k (y′), 1≤y′≤I _k ′×J _k ′}, where I _k ′ represents the vertical resolution of the kth binary digital watermark W′ _k to be extracted, and J _k ′ represents The horizontal resolution of the kth binary digital watermark W′ _k to be extracted, tfacs _k (y′) represents the y′ discrete cosine transform AC coefficient in TFACS _k ;

②-7. Compare the size of any two adjacent discrete cosine transform AC coefficients in TFACS ₁ , TFACS ₂ , ..., TFACS _k , ... and TFACS _K respectively, and return a logical value 1 or 0 according to the comparison result. For TFACS _k Any two adjacent discrete cosine transform AC coefficients in , record them as tfacs _k (z′) and tfacs _k (z′+1) respectively, judge tfacs _k (z′)>tfacs _k (z′+1) Whether it is true, if true, return a logical value 1, otherwise, return a logical value 0, where, 1≤z′≤I _k ′×J _k ′-1; then compare TFACS ₁ , TFACS ₂ , ..., TFACS _k , ... and the size of the last DCT AC coefficient and the first DCT coefficient in TFACS _K , and return a logical value 1 or 0 according to the comparison result, for the last DCT AC coefficient and the first DCT coefficient in TFACS _k A discrete cosine transform AC coefficient, if the _{former is} large, return a logical value of 1, otherwise, return _a logical value _of 0; then construct One-to-one correspondence with the second binary digital watermark key, for the returned logical value corresponding to TFACS _k , store the returned logical value in a two-dimensional array with a size of I _k ′×J _k ′ In the matrix, the two-dimensional matrix is used as the kth second binary digital watermark key, denoted as TWB _k ;

②-8. Combine K pieces of zero-watermark information WO ₁ , WO ₂ , ..., WO _k , ... and WO _K from multiple zero-watermark embedding terminals with K second binary digital watermark keys TWB ₁ , TWB ₂ respectively , ..., TWB _k , ... and TWB _K are one-to-one XOR operation, and K binary digital watermarks are recovered, which are respectively recorded as TW ₁ , TW ₂ , ..., TW _k , ... and TW _K , TW ₁ = xor(WO ₁ ,TWB ₁ ), TW ₂ =xor(WO ₂ ,TWB ₂ ),…,TW _k =xor(WO _k ,TWB _k ),…,TW _K =xor(WO _K ,TWB _K ), where , TWB ₁ represents the first second binary digital watermark key, TWB ₂ represents the second second binary digital watermark key, TWB _k represents the k-th second binary digital watermark key, TWB _K represents the second K second binary digital watermark keys, xor () is an exclusive OR operation function;

和 ②-9. Perform anti-scrambling processing on K binary digital watermarks TW ₁ , TW ₂ , ..., TW _k , ... and TW _K respectively, to obtain K binary digital watermarks with copyright authentication information, which are respectively recorded as

and

和

分别与多个零水印嵌入端嵌入的K个二值数字水印W₁、W₂、…、W_k、…和W_K一一对应进行相似度计算，对应得到K个归一化相关系数，然后根据K个归一化相关系数的大小判定是否提取出多个零水印嵌入端嵌入的K个二值数字水印W₁、W₂、…、W_k、…和W_K。②-10. K binary digital watermarks with copyright authentication information

and

Carry out similarity calculation in one-to-one correspondence with K binary digital watermarks W ₁ , W ₂ , ..., W _k , ... and W _K embedded in multiple zero-watermark embedding terminals, and obtain K normalized correlation coefficients correspondingly, and then Determine whether to extract K binary digital watermarks W ₁ , W ₂ , . . . , W _{k ,} .

The concrete process of described step ②-10 is:

和分别与多个零水印嵌入端嵌入的K个二值数字水印W₁、W₂、…、W_k、…和W_K一一对应进行相似度计算，对应得到K个归一化相关系数，对于将

与W_k进行相似度计算后得到的第k个归一化相关系数，将其记为

$\mathrm{\ρ}(W_k,W_k^{*})=\frac{\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}((w_k(i_k,j_k)-\stackrel{\‾}{w_k})\×(w_k^{*}(i_k,j_k)-\stackrel{\‾}{w_k^{*}}))}{\sqrt{\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}{(w_k(i_k,j_k)-\stackrel{\‾}{w_k})}^2\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}{(w_k^{*}(i_k,j_k)-\stackrel{\‾}{w_k^{*}})}^2}},$ 其中，w_k(i_k,j_k)表示多个零水印嵌入端嵌入的第k个二值数字水印W_k中坐标位置为(i_k,j_k)的像素的像素值，多个零水印嵌入端嵌入的第k个二值数字水印W_k中所有像素的像素值的均值，

表示具有版权认证信息的第k个二值数字水印

中坐标位置为(i_k,j_k)的像素的像素值，

表示具有版权认证信息的第k个二值数字水印

中所有像素的像素值的均值；z1. K binary digital watermarks with copyright authentication information

and Carry out one-to-one similarity calculation with the K binary digital watermarks W ₁ , W ₂ , ..., W _k , ... and W _K embedded in multiple zero-watermark embedding terminals, and obtain K normalized correlation coefficients correspondingly. For Will

The kth normalized correlation coefficient obtained after similarity calculation with W _k is recorded as

$\mathrm{\ρ}(W_k,W_k^{*})=\frac{\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}((w_k(i_k,j_k)-\stackrel{\‾}{w_k})\×(w_k^{*}(i_k,j_k)-\stackrel{\‾}{w_k^{*}}))}{\sqrt{\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}{(w_k(i_k,j_k)-\stackrel{\‾}{w_k})}^2\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}{(w_k^{*}(i_k,j_k)-\stackrel{\‾}{w_k^{*}})}^2}},$ Among them, w _k (i _k , j _k ) represents the pixel value of the pixel whose coordinate position is (i _k , j _k ) in the kth binary digital watermark W _k embedded by multiple zero-watermark embedding terminals, The mean value of the pixel values of all pixels in the kth binary digital watermark W _k embedded in multiple zero watermark embedding terminals,

Represents the kth binary digital watermark with copyright authentication information

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

Represents the kth binary digital watermark with copyright authentication information

The mean of the pixel values of all pixels in ;

如果

的值为1，则确定W_k被无损地提取出，如果

的值大于或等于δ_T并小于1，则确定W_k提取成功，如果

的值小于δ_T，则确定W_k提取失败，其中，δ_T表示水印提取门限。z2. Determine whether to extract K binary digital watermarks W ₁ , W ₂ , ..., W _k , ... and W _K embedded in multiple zero-watermark embedding terminals according to the size of the K normalized correlation coefficients. correlation coefficient

if

A value of 1 determines that W _k is extracted losslessly, if

The value of is greater than or equal to δ _T and less than 1, then it is determined that the extraction of W _k is successful, if

If the value of is less than δ _T , it is determined that W _k extraction fails, where δ _T represents the watermark extraction threshold.

Compared with the prior art, the present invention has the advantages of:

1) Compared with the existing single zero watermark technology, the embedding method of the present invention can realize the simultaneous embedding of multiple binary digital watermarks for different purposes and purposes according to the needs of practical applications, and the embedded multiple binary digital watermarks Watermarks do not interfere with each other, so it can better meet the application diversity of digital watermarks, and has greater practical application value and broad application prospects.

2) Compared with the traditional digital watermarking technology, the embedding method of the present invention does not embed multiple binary digital watermarks into the original digital image, but registers them in the digital watermark information database, so not only embedding The binary digital watermark is completely imperceptible, and does not cause any damage to the original digital image data, there is no problem of image quality degradation, and the integrity of the original digital image information is maintained, which is very suitable for some sensitive digital images such as Copyright protection in areas such as medical images, palm print images, and military images.

3) The embedding and extraction method of the present invention is carried out in the composite domain of two-dimensional discrete wavelet transform (DWT) and two-dimensional discrete cosine transform (DCT), making full use of the characteristics of both, thus achieving excellent robust performance The embedding and extraction of multiple zero watermarks against conventional image processing attacks well balances the contradiction between the robustness and imperceptibility of digital watermarking.

Description of drawings

Figure 1a is the original Lena digital image;

Figure 1b is the original binary signature watermark;

Figure 1c is the original binary serial number watermark;

Figure 1d is the original binary icon watermark;

Figure 2a is a Lena digital image embedded with a binary watermark;

Figure 2b is the Lena digital image embedded with two binary watermarks;

Figure 2c is a Lena digital image embedded with three binary watermarks;

Figure 2d is the binary signature watermark extracted from Figure 2c;

Fig. 2e is the binary serial number watermark extracted from Fig. 2c;

Figure 2f is the binary icon watermark extracted from Figure 2c;

Fig. 3 a is the watermark Lena digital image after all pixel values are added 0.2 to the digital image shown in Fig. 2c;

Fig. 3b is the watermark Lena digital image after adding 0.5 to all pixel values of the digital image shown in Fig. 2c;

Fig. 3c is the watermark Lena digital image after all pixel values are subtracted by 0.2 to the digital image shown in Fig. 2c;

Figure 3d is the watermarked Lena digital image after subtracting 0.5 from all pixel values of the digital image shown in Figure 2c;

Fig. 3e is the binary signature watermark extracted from Fig. 3a, Fig. 3b and Fig. 3c;

Fig. 3f is the binary serial number watermark extracted from Fig. 3a, Fig. 3b and Fig. 3c;

Fig. 3g is the binary icon watermark extracted from Fig. 3a, Fig. 3b and Fig. 3c;

Fig. 4a is the watermarked Lena digital image after histogram equalization processing to the digital image shown in Fig. 2c;

Fig. 4b is the binary signature watermark extracted from Fig. 4a;

Fig. 4c is the binary serial number watermark extracted from Fig. 4a;

Figure 4d is a binary icon watermark extracted from Figure 4a;

Fig. 5a is the watermarked Lena digital image after the digital image shown in Fig. 2c is filtered by [5 * 5] window median;

Fig. 5b is the watermarked Lena digital image after the digital image shown in Fig. 2c is filtered by [11×11] window median;

Figure 5c is the binary signature watermark extracted from Figure 5a;

Figure 5d is the binary serial number watermark extracted from Figure 5a;

Figure 5e is a binary icon watermark extracted from Figure 5a;

Figure 5f is the binary signature watermark extracted from Figure 5b;

Figure 5g is the binary serial number watermark extracted from Figure 5b;

Figure 5h is a binary icon watermark extracted from Figure 5b;

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

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

Fig. 6c is the binary signature watermark extracted from Fig. 6a;

Figure 6d is the binary serial number watermark extracted from Figure 6a;

Figure 6e is a binary icon watermark extracted from Figure 6a;

Figure 6f is the binary signature watermark extracted from Figure 6b;

Fig. 6g is the binary serial number watermark extracted from Fig. 6b;

Figure 6h is the binary icon watermark extracted from Figure 6b;

Figure 7a is the watermarked Lena digital image after superimposing the digital image shown in Figure 2c with a mean value of 0 and a variance of 0.02 Gaussian noise;

Figure 7b is the watermarked Lena digital image after superimposing the digital image shown in Figure 2c with a mean value of 0 and a variance of 0.05 Gaussian noise;

Figure 7c is the binary signature watermark extracted from Figure 7a;

Figure 7d is the binary serial number watermark extracted from Figure 7a;

Figure 7e is a binary icon watermark extracted from Figure 7a;

Figure 7f is the binary signature watermark extracted from Figure 7b;

Figure 7g is the binary serial number watermark extracted from Figure 7b;

Figure 7h is a binary icon watermark extracted from Figure 7b;

Fig. 8a is the watermark Lena digital image after cutting the upper left corner 128 * 128 pixels from the digital image shown in Fig. 2c;

Fig. 8b is the watermarked Lena digital image after cutting the upper left corner 256×256 pixels from the digital image shown in Fig. 2c;

Figure 8c is the binary signature watermark extracted from Figure 8a;

Figure 8d is the binary serial number watermark extracted from Figure 8a;

Figure 8e is a binary icon watermark extracted from Figure 8a;

Figure 8f is the binary signature watermark extracted from Figure 8b;

Figure 8g is the binary serial number watermark extracted from Figure 8b;

Figure 8h is a binary icon watermark extracted from Figure 8b;

Figure 9a is the watermarked Lena digital image after the digital image shown in Figure 2c is rotated 5 degrees counterclockwise and then restored to its original direction;

Figure 9b is the watermarked Lena digital image after the digital image shown in Figure 2c is rotated 25 degrees counterclockwise and then restored to its original direction;

Figure 9c is the binary signature watermark extracted from Figure 9a;

Figure 9d is the binary serial number watermark extracted from Figure 9a;

Figure 9e is a binary icon watermark extracted from Figure 9a;

Figure 9f is the binary signature watermark extracted from Figure 9b;

Figure 9g is the binary serial number watermark extracted from Figure 9b;

Figure 9h is the binary icon watermark extracted from Figure 9b.

Detailed ways

The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments.

The invention proposes a method for embedding and extracting multiple zero watermarks of a digital image. The watermark embedding and extraction are all carried out in the compound domain of discrete wavelet transform and discrete cosine transform, and the excellent characteristics of both are fully utilized. The main process of multiple zero-watermark embedding is as follows: firstly, perform discrete wavelet transform of appropriate levels on the original digital image, and then perform discrete cosine transform on the obtained wavelet approximation sub-graph, and then according to the resolution of the K binary digital watermarks to be embedded rate, respectively select K groups of DCT AC coefficients that meet the absolute value requirements of a certain difference to form K discrete cosine transform AC coefficient sequences, and finally construct K according to the relationship between the discrete cosine transform AC coefficients in each sequence. A robust binary digital watermark key, which is XORed with K practically meaningful binary digital watermarks to be embedded, and then saved to the registration agency to complete the embedding of K zero-watermarks; the main method for extracting multiple zero-watermarks The process is: through the similar discrete wavelet transform and discrete cosine transform processing process of the digital image to be extracted K binary digital watermarks, combined with the K zero watermark information stored in the registration agency, extract K related watermarks to prove the Copyright or ownership of the original digital image.

A kind of multiple zero watermark embedding method of digital image of the present invention, it comprises the following steps:

①-1. At multiple zero-watermark embedding terminals, assume that the original digital image to be embedded with K binary digital watermarks is an 8-bit grayscale image, and denoted as F, F={0≤f(m,n)≤255 ,1≤m≤M,1≤n≤N}, where K≥2, M represents the vertical resolution of the original digital image F to be embedded with K binary digital watermarks, and N represents the K binary digital watermarks to be embedded The horizontal resolution of the original digital image F of the watermark, M×N represents the resolution of the original digital image F to be embedded with K binary digital watermarks, f(m,n) represents the original digital image F to be embedded with K binary digital watermarks The pixel value of the pixel whose coordinate position is (m, n) in the image F.

①-2. At multiple zero-watermark embedding terminals, assuming that the K digital watermarks to be embedded are all binary images, which are recorded as W ₁ , W ₂ ,...,W _k ,...,W _K , then for the The k-th binary digital watermark W _k , W _k ={w _k (i _k ,j _k )=0 or 1,1≤i _k ≤I _k ,1≤j _k ≤J _k }, where, 1≤ k≤K, I _k represents the vertical resolution of the k-th binary digital watermark W _k to be embedded, J _k represents the horizontal resolution of the k-th binary digital watermark W _k to be embedded, I _k ×J _k Indicates the resolution of the kth binary digital watermark W _k to be embedded, w _k (i _k , j _k ) indicates the coordinate position of the kth binary digital watermark W _k to be embedded is (i _k , j _k ) The pixel value of the pixel.

In this specific embodiment, the K digital watermarks to be embedded can be binary images such as signatures, logos, serial numbers, dates or company icons of authors or owners of corresponding copyright information or ownership information with practical significance.

①-3. The original digital image F to be embedded with K binary digital watermarks is normalized to obtain the normalized digital image, denoted as F′, and the normalized digital image F′ is The pixel value of the pixel whose coordinate position is (m,n) is recorded as f'(m,n), f'(m,n)=f(m,n)/255.

min()为取最小值函数，max()为取最大值函数，符号表示取小于其自身的最大整数，即该符号为向下取整符号，I₁表示待嵌入的第1个二值数字水印W₁的竖直分辨率，J₁表示待嵌入的第1个二值数字水印W₁的横向分辨率，I_K表示待嵌入的第K个二值数字水印W_K的竖直分辨率，J_K表示待嵌入的第K个二值数字水印W_K的横向分辨率。①-4. Carry out L-level two-dimensional discrete wavelet transform to F′ to obtain a first wavelet approximation subgraph and multiple first wavelet detail subgraphs, and denote the first wavelet approximation subgraph as FA, where the resolution of FA The rate is (M/2 ^L )×(N/2 ^L ),

min() is the minimum value function, max() is the maximum value function, the symbol Indicates to take the largest integer smaller than itself, that is, the symbol is a rounded-down symbol, I ₁ represents the vertical resolution of the first binary digital watermark W ₁ to be embedded, J ₁ represents the first binary digital watermark to be embedded The horizontal resolution of the value digital watermark W ₁ , I _K represents the vertical resolution of the Kth binary digital watermark W _K to be embedded, and J _K represents the horizontal resolution of the Kth binary digital watermark W _K to be embedded .

①-5. Carry out two-dimensional discrete cosine transform to FA to obtain a first two-dimensional discrete cosine transform coefficient matrix with the same resolution as FA, denoted as FAC, and then carry out Zig-Zag scanning arrangement on FAC to obtain a first one dimensional discrete cosine transform coefficient sequence, denoted as FACS, FACS={facs(x),1≤x≤(M/2 ^L )×(N/2 ^L )}, where facs(x) represents the x-th in FACS The first discrete cosine transform coefficient in FACS is the first discrete cosine transform AC coefficient.

①-6. According to the resolution of each binary digital watermark to be embedded, select I ₁ ×J ₁ , I ₂ ×J ₂ ,…, I _k ×J _k , ...and I _K ×J _K discrete cosine transform AC coefficients to form K first one-dimensional discrete cosine transform AC coefficient sequences, respectively denoted as FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K , for from The kth first one-dimensional discrete cosine transform AC coefficient sequence FACS k composed of I _k ×J _k discrete cosine transform AC coefficients selected in FACS that _{meet the set conditions, FACS k =} {facs _k (y),1 ≤y≤I _k ×J _k }, and then record the corresponding position information of each discrete cosine transform AC coefficient in FACS in FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K , where I ₂ represents the The vertical resolution of the second binary digital watermark W ₂ , J ₂ represents the horizontal resolution of the second binary digital watermark W ₂ to be embedded, and facs _k (y) represents the yth discrete value in FACS _k For the cosine transform AC coefficient, the setting condition is such that the absolute value of the difference between any two adjacent discrete cosine transform AC coefficients in the formed first one-dimensional discrete cosine transform AC coefficient sequence is greater than or equal to the set difference threshold.

Here, the selection of the discrete cosine transform AC coefficients in each first one-dimensional discrete cosine transform AC coefficient sequence starts from the second discrete cosine transform coefficient in the first one-dimensional discrete cosine transform coefficient sequence FACS (that is, the second discrete cosine transform coefficient 1 discrete cosine transform AC coefficient), and any two adjacent discrete cosines in the selected K first one-dimensional discrete cosine transform AC coefficient sequences FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K The absolute values of the differences of the transformed AC coefficients are respectively greater than or equal to δ ₁ , δ ₂ , ..., δ _k , ... and δ _K , where δ ₁ represents the difference between any two adjacent discrete cosine transform AC coefficients in FACS ₁ The first difference threshold set by the absolute value of the difference, δ ₂ represents the second difference threshold set for the absolute value of the difference between any two adjacent discrete cosine transform AC coefficients in FACS ₂ , δ _k represents the kth difference threshold set for the absolute value of the difference between any two adjacent discrete cosine transform AC coefficients in FACS _k , and δ _K represents the value for any adjacent two discrete cosine transform AC coefficients in FACS _K The absolute value of the coefficient difference sets the Kth difference threshold. Taking δ _k as an example, the setting principle of δ _k is as follows: take the maximum value under the premise of ensuring that I _k × J _k discrete cosine transform AC coefficients can be selected, so as to improve the selection of digital images when they are under attack. The invariance of the relationship between the magnitudes of these discrete cosine transform AC coefficients.

①-7. Compare the size of any two adjacent discrete cosine transform AC coefficients in FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K respectively, and return a logical value 1 or 0 according to the comparison result, for FACS _k Any two adjacent discrete cosine transform AC coefficients in , record them as facs _k (z) and facs _k (z+1) respectively, and judge whether facs _k (z)>facs _k (z+1) is true, if If it is true, return a logic value of 1, otherwise, return a logic value of 0, where, 1≤z≤I _k ×J _k -1; then compare the last one of FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K DCT AC coefficient and the size of the first DCT AC coefficient, and return a logic value of 1 or 0 according to the comparison result. For the last DCT AC coefficient and the first DCT AC coefficient in FACS _k , If _the former is greater, return _a logical value of 1, otherwise, return a logical value of ₀ ; then construct a one-to-one correspondence of the first and _second value digital watermark key, for the returned logical value corresponding to FACS _k , the returned logical value is stored in a two-dimensional matrix with a size of I _k × J _k in the order of first row and second column, and the two-dimensional matrix is used as The k-th first binary digital watermark key is denoted as WB _k .

①-8. The K binary digital watermarks W ₁ , W ₂ ,...,W _k ,...,W _K to be embedded are respectively scrambled, and the K binary digital watermarks obtained after the scrambled processing are respectively corresponding to are WS ₁ , WS ₂ , ..., WS _k , ... and _WSK , then combine WS ₁ , WS ₂ , ..., WS _k , ... and WS _K with K first binary digital watermark keys WB ₁ , WB ₂ , ..., WB _k , ... and WB _K are one-to-one XORed to obtain K zero-watermark information, respectively recorded as WO ₁ , WO ₂ , ..., WO _k , ... and WO _K , WO ₁ =xor( WS ₁ ,WB ₁ ), WO ₂ =xor(WS ₂ ,WB ₂ ),…, WO _k =xor(WS _k ,WB _k ),…, WO _K =xor(WS _K ,WB _K ), and K Zero watermark information WO ₁ , WO ₂ , ..., WO _k , ... and WO _K are stored in the digital watermark information database of the registration authority, and K binary digital watermarks W ₁ , W ₂ , ..., W _k , ..., The embedding of W _K , where WS ₁ represents the binary digital watermark obtained after scrambling W ₁ , WS ₂ represents the binary digital watermark obtained after scrambling W ₂ , and WS _k represents the binary watermark obtained after scrambling W _k The binary digital watermark obtained after scrambling processing, WS _K represents the binary digital watermark obtained after scrambling W _K , WB ₁ represents the first first binary digital watermark key, WB ₂ represents the second The first binary digital watermark key, WB _K represents the Kth first binary digital watermark key, and xor() is an exclusive OR operation function.

Here, scrambling the K binary digital watermarks W ₁ , W ₂ ,...,W _k ,...,W _K to be embedded can effectively improve the security of the zero watermark in the registration authority. Here, the scrambling The processing can use conventional Arnold transformation, or other existing mature scrambling methods.

①-9. At multiple zero-watermark embedding terminals, record the position information of each discrete cosine transform AC coefficient in FACS in FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K , and K zero-watermark information WO ₁ , WO ₂ ,..., WO _k ,..., WO _K and K binary digital watermarks W ₁ , W ₂ ,..., W _k ,..., W _K are transmitted to multiple zero-watermark extraction terminals.

A method for extracting a plurality of zero watermarks of a digital image of the present invention comprises the following steps:

②-1. At multiple zero-watermark extraction terminals, record the digital images (that is, test images) of K binary digital watermarks to be extracted as TF, TF={0≤tf(m′,n′)≤255,1 ≤m'≤M',1≤n'≤N'}, where K≥2, M' represents the vertical resolution of the digital image TF of K binary digital watermarks to be extracted, and N' represents the K binary watermarks to be extracted The horizontal resolution of the digital image TF of the binary digital watermark, M'×N' represents the resolution of the digital image TF of K binary digital watermarks to be extracted, and the resolution of the digital image TF of K binary digital watermarks to be extracted The resolution of digital images embedded with digital watermarks at multiple zero-watermark embedding terminals is the same, and tf(m',n') means that the coordinate positions in the digital image TF of K binary digital watermarks to be extracted are (m',n' ) pixel value of the pixel.

②-2. At multiple zero-watermark extraction terminals, mark the K binary digital watermarks to be extracted as W′ ₁ , W′ ₂ , ..., W′ _k , ... and W′ _K . k binary digital watermarks W′ _k , W′ _k ={w′ _k (i′ _k ,j′ _k )=0 or 1, 1≤i′ _k ≤I′ _k ,1≤j′ _k ≤J′ _k }, where, 1≤k≤K, I′ _k represents the vertical resolution of the kth binary digital watermark W′ _k to be extracted, and J′ _k represents the kth binary digital watermark W′ to be extracted The horizontal resolution of _k , I′ _k ×J′ _k represents the resolution of the kth binary digital watermark W′ _k to be extracted, and the resolution of the kth binary digital watermark W′ _k to be extracted is related to multiple The resolution of the kth binary digital watermark embedded in the zero watermark embedding end is the same, w′ _k (i′ _k , j′ _k ) means that the coordinate position of the kth binary digital watermark W′ _k to be extracted is (i ′ _k , j′ _k ) pixel value of the pixel.

②-3. Perform normalization processing on the digital image TF to be extracted K binary digital watermarks, and obtain the digital image after normalization processing, which is denoted as TF′, and the coordinates in the digital image TF′ after normalization processing The pixel value of the pixel at the position (m',n') is recorded as tf'(m',n'), tf'(m',n')=tf(m',n')/255.

min()为取最小值函数，max()为取最大值函数，符号

表示取小于其自身的最大整数，即该符号为向下取整符号，I₁′表示待提取的第1个二值数字水印W′₁的竖直分辨率，J₁′表示待提取的第1个二值数字水印W′₁的横向分辨率，I_K′表示待提取的第K个二值数字水印W′_K的竖直分辨率，J_K′表示待提取的第K个二值数字水印W′_K的横向分辨率。②-4. Carry out L'-level two-dimensional discrete wavelet transform to TF' to obtain a second wavelet approximation subgraph and a plurality of second wavelet detail subgraphs, and denote the second wavelet approximation subgraph as TFA, wherein the TFA's The resolution is (M′/2 ^L ′)×(N′/2 ^L ′),

min() is the minimum value function, max() is the maximum value function, the symbol

Indicates to take the largest integer smaller than itself, that is, the symbol is a rounded-down symbol, I ₁ ′ represents the vertical resolution of the first binary digital watermark W′ ₁ to be extracted, and J ₁ ′ represents the first binary watermark to be extracted 1 horizontal resolution of a binary digital watermark W′ ₁ , I _K ′ represents the vertical resolution of the Kth binary digital watermark W′ _K to be extracted, and J _K ′ represents the Kth binary number to be extracted The lateral resolution of the watermark W′ _K.

②-5. Perform two-dimensional discrete cosine transform on TFA to obtain a second two-dimensional discrete cosine transform coefficient matrix with the same resolution as TFA, denoted as TFAC, and then carry out Zig-Zag scanning arrangement on TFAC to obtain a second one dimensional discrete cosine transform coefficient sequence, denoted as TFACS, TFACS={tfacs(x′),1≤x′≤(M′/2 ^L ′)×(N′/2 ^L ′)}, where, tfacs(x′ ) represents the x′th discrete cosine transform coefficient in TFACS, and the second discrete cosine transform coefficient in TFACS is initially an AC coefficient of discrete cosine transform.

②-6. According to the corresponding position information in FACS of each discrete cosine transform AC coefficient in FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K recorded by multiple zero-watermark embedding terminals, extract from TFACS respectively The I ₁ ′×J ₁ ′, I ₂ ′×J ₂ ′, ..., I _k ′×J _k ′, ... and I _K ′×J _K ′ discrete cosine transform AC coefficients at the corresponding positions constitute K second The one-dimensional discrete cosine transform AC coefficient sequence is correspondingly denoted as TFACS ₁ , TFACS ₂ , ..., TFACS _k , ... and TFACS _K , for each discrete cosine transform AC coefficient in FACS _k recorded according to multiple zero-watermark embedding terminals in Corresponding position information in FACS, extract the kth second one-dimensional discrete cosine transform AC coefficient sequence TFACS k composed of I _k ′×J _k ′ discrete cosine transform AC coefficients corresponding to the position from TFACS, TFACS _{k =} { tfacs _k (y′), 1≤y′≤I _k ′×J _k ′}, where I _k ′ represents the vertical resolution of the kth binary digital watermark W′ _k to be extracted, and J _k ′ represents The horizontal resolution of the kth binary digital watermark W′ _k to be extracted, tfacs _k (y′) represents the y′th discrete cosine transform AC coefficient in TFACS _k .

②-7. Compare the size of any two adjacent discrete cosine transform AC coefficients in TFACS ₁ , TFACS ₂ , ..., TFACS _k , ... and TFACS _K respectively, and return a logical value 1 or 0 according to the comparison result. For TFACS _k Any two adjacent discrete cosine transform AC coefficients in , record them as tfacs _k (z′) and tfacs _k (z′+1) respectively, judge tfacs _k (z′)>tfacs _k (z′+1) Whether it is true, if true, return a logical value 1, otherwise, return a logical value 0, where, 1≤z′≤I _k ′×J _k ′-1; then compare TFACS ₁ , TFACS ₂ , ..., TFACS _k , ... and the size of the last DCT AC coefficient and the first DCT coefficient in TFACS _K , and return a logical value 1 or 0 according to the comparison result, for the last DCT AC coefficient and the first DCT coefficient in TFACS _k A discrete cosine transform AC coefficient, if the _former is _large , return a logical value of 1, otherwise, return _a logical value of ₀ ; then construct One-to-one correspondence with the second binary digital watermark key, for the returned logical value corresponding to TFACS _k , store the returned logical value in a two-dimensional array with a size of I _k ′×J _k ′ In the matrix, the two-dimensional matrix is used as the kth second binary digital watermark key, which is denoted as TWB _k .

②-8. Combine K pieces of zero-watermark information WO ₁ , WO ₂ , ..., WO _k , ... and WO _K from multiple zero-watermark embedding terminals with K second binary digital watermark keys TWB ₁ , TWB ₂ respectively , ..., TWB _k , ... and TWB _K are one-to-one XOR operation, and K binary digital watermarks are recovered, which are respectively recorded as TW ₁ , TW ₂ , ..., TW _k , ... and TW _K , TW ₁ = xor(WO ₁ ,TWB ₁ ), TW ₂ =xor(WO ₂ ,TWB ₂ ),...,TW _k =xor(WO _k ,TWB _k ),...,TW _K =xor(WO _K ,TWB _K ), where , TWB ₁ represents the first second binary digital watermark key, TWB ₂ represents the second second binary digital watermark key, TWB _k represents the k-th second binary digital watermark key, TWB _K represents the second K second binary digital watermark keys, and xor() is an exclusive OR operation function.

和

②-9. Perform anti-scrambling processing on K binary digital watermarks TW ₁ , TW ₂ , ..., TW _k , ... and TW _K respectively, to obtain K binary digital watermarks with copyright authentication information, which are respectively recorded as

and

和

分别与多个零水印嵌入端嵌入的K个二值数字水印W₁、W₂、…、W_k、…和W_K一一对应进行相似度计算，对应得到K个归一化相关系数，然后根据K个归一化相关系数的大小来判定是否提取出多个零水印嵌入端嵌入的K个二值数字水印W₁、W₂、…、W_k、…和W_K以及嵌入的K个二值数字水印的鲁棒性。②-10. K binary digital watermarks with copyright authentication information

and

Carry out similarity calculation in one-to-one correspondence with K binary digital watermarks W ₁ , W ₂ , ..., W _k , ... and W _K embedded in multiple zero-watermark embedding terminals, and obtain K normalized correlation coefficients correspondingly, and then According to the size of the K normalized correlation coefficients, it is judged whether to extract the K binary digital watermarks W ₁ , W ₂ ,..., W _k ,... and W _K embedded in multiple zero-watermark embedding terminals and the embedded K binary Robustness of digital watermarking.

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

和分别与多个零水印嵌入端嵌入的K个二值数字水印W₁、W₂、…、W_k、…和W_K一一对应进行相似度计算，对应得到K个归一化相关系数，对于将

与W_k进行相似度计算后得到的第k个归一化相关系数，将其记为

$\mathrm{\ρ}(W_k,W_k^{*})=\frac{\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}((w_k(i_k,j_k)-\stackrel{\‾}{w_k})\×(w_k^{*}(i_k,j_k)-\stackrel{\‾}{w_k^{*}}))}{\sqrt{\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}{(w_k(i_k,j_k)-\stackrel{\‾}{w_k})}^2\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}{(w_k^{*}(i_k,j_k)-\stackrel{\‾}{w_k^{*}})}^2}},$ 其中，w_k(i_k,j_k)表示多个零水印嵌入端嵌入的第k个二值数字水印W_k中坐标位置为(i_k,j_k)的像素的像素值，

表示多个零水印嵌入端嵌入的第k个二值数字水印W_k中所有像素的像素值的均值，

表示具有版权认证信息的第k个二值数字水印

中坐标位置为(i_k,j_k)的像素的像素值，

表示具有版权认证信息的第k个二值数字水印

中所有像素的像素值的均值。z1. K binary digital watermarks with copyright authentication information

and Carry out one-to-one similarity calculation with the K binary digital watermarks W ₁ , W ₂ , ..., W _k , ... and W _K embedded in multiple zero-watermark embedding terminals, and obtain K normalized correlation coefficients correspondingly. For Will

The kth normalized correlation coefficient obtained after similarity calculation with W _k is recorded as

$\mathrm{\ρ}(W_k,W_k^{*})=\frac{\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}((w_k(i_k,j_k)-\stackrel{\‾}{w_k})\×(w_k^{*}(i_k,j_k)-\stackrel{\‾}{w_k^{*}}))}{\sqrt{\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}{(w_k(i_k,j_k)-\stackrel{\‾}{w_k})}^2\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}{(w_k^{*}(i_k,j_k)-\stackrel{\‾}{w_k^{*}})}^2}},$ Among them, w _k (i _k , j _k ) represents the pixel value of the pixel whose coordinate position is (i _k , j _k ) in the kth binary digital watermark W _k embedded by multiple zero-watermark embedding terminals,

Represents the mean value of the pixel values of all pixels in the kth binary digital watermark W _k embedded in multiple zero-watermark embedding terminals,

Represents the kth binary digital watermark with copyright authentication information

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

Represents the kth binary digital watermark with copyright authentication information

The mean of the pixel values of all pixels in .

判断

的值是否为1，如果的值为1，则确定与W_k完全一致，表明W_k被无损地提取出，如果这是嵌入水印后的Lena数字图像受到某种处理或攻击后的提取结果，说明嵌入的二值数字水印W_k具有抵抗这种处理或攻击的能力，具有理想的鲁棒性。如果

的值不为1，则再判断

的值是否介于δ_T-1，如果

的值介于δ_T-1，则说明提取出的二值数字水印

与嵌入端嵌入的二值数字水印W_k存在一定差异，但两者间具有很大的相似性，这时可以从提取结果中辨认出嵌入的二值数字水印W_k，提取成功。相关系数的值越大，提取出的二值数字水印

与嵌入端嵌入的二值数字水印W_k越相似，越容易辨认出嵌入的二值数字水印W_k，提取效果越好。如果这是嵌入水印后的Lena数字图像受到某种处理或攻击后的提取结果，说明嵌入的二值数字水印W_k具有比较理想的抵抗这种处理或攻击的能力，具有很好的鲁棒性。其中，δ_T为水印提取门限，一般可取值为0.5。如果

的值不介于δ_T-1，则再判断

的值是否小于δ_T，如果是，则说明提取出的二值数字水印

与嵌入端嵌入的二值数字水印W_k相关性很小，这时已无法从提取结果中辨认出嵌入的二值数字水印W_k，提取失败。z2. Determine whether to extract K binary digital watermarks W ₁ , W ₂ , ..., W _k , ... and W _K embedded in multiple zero-watermark embedding terminals according to the size of the K normalized correlation coefficients. correlation coefficient

judge

Is the value of 1, if value of 1, then determine It is completely consistent with W _k , indicating that W _k is extracted losslessly. If this is the extraction result after the watermarked Lena digital image is subjected to some kind of processing or attack, it means that the embedded binary digital watermark W _k has the ability to resist this processing. Or the ability to attack, with ideal robustness. if

value is not 1, then judge again

Whether the value of is between δ _T -1, if

If the value is between δ _T -1, it means that the extracted binary digital watermark

There are certain differences with the binary digital watermark W _k embedded in the embedding terminal, but there is a great similarity between them. At this time, the embedded binary digital watermark W _k can be identified from the extraction result, and the extraction is successful. The larger the value of the correlation coefficient, the extracted binary digital watermark

The more similar it is to the binary digital watermark W _k embedded in the embedding end, the easier it is to identify the embedded binary digital watermark W _k , and the better the extraction effect is. If this is the extraction result of the embedded watermarked Lena digital image after some kind of processing or attack, it shows that the embedded binary digital watermark W _k has an ideal ability to resist such processing or attack, and has good robustness . Among them, δ _T is the threshold of watermark extraction, and the generally desirable value is 0.5. if

value is not between δ _T -1, then judge

Is the value of is less than δ _T , if yes, it means that the extracted binary digital watermark

The correlation with the binary digital watermark W _k embedded in the embedding terminal is very small, and the embedded binary digital watermark W _k cannot be recognized from the extraction result at this time, and the extraction fails.

In order to better illustrate the feasibility and effectiveness of a multiple zero-watermark embedding and extraction method for a digital image proposed by the present invention, taking embedding three binary digital watermarks as an example (K=3), the simulation is carried out through the following experiments .

The experimental simulation is carried out on the Matlab7.5 platform. The original digital image F to be embedded with three binary digital watermarks is an 8-bit Lena grayscale image with a resolution of 512×512, as shown in Figure 1a. The three binary digital watermarks are binary signature watermark W ₁ , binary serial number watermark W ₂ and binary icon watermark W ₃ , with resolutions of 28×50, 16×64 and 25×32, respectively, as shown in Fig. 1b, Figure 1c and Figure 1d show.

Before the binary digital watermark is embedded, normalize the pixels of the original digital image F, and then perform three-level two-dimensional discrete wavelet transform on the normalized digital image F'.

The quality of the digital image embedded with the digital watermark is judged by the peak signal-to-noise ratio (PSNR). $\mathrm{PSNR}=-10\×\log 10\left(\frac{\underset{m=1}{\overset{m}{\mathrm{\Σ}}}\underset{\mathrm{no}=1}{\overset{N}{\mathrm{\Σ}}}{(f(m,\mathrm{no})-f^{\′\′}(m,\mathrm{no}))}^2}{m\×N\×f_{\max }^2}\right),$ Among them, f(m,n) represents the pixel value of the pixel whose coordinate position is (m,n) in the original digital image to be embedded with binary digital watermark, and f″(m,n) represents the number after embedding binary digital watermark The pixel value of the pixel whose coordinate position is (m, n) in the image, f _max represents the maximum pixel value of the original digital image.

表示提取出的第k个二值数字水印

中坐标位置为(i_k,j_k)的像素的像素值，和

分别表示数字水印W_k和

的所有像素的像素值的均值。根据归一化相关系数ρ的大小可以判定是否提取出嵌入的二值数字水印以及嵌入的二值数字水印鲁棒性如何。The objective evaluation of digital watermark detection results uses the normalized correlation coefficient (ρ): $\mathrm{\ρ}(W_k,W_k^{*})=\frac{\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}((w_k(i_k,j_k)-\stackrel{\‾}{w_k})\×(w_k^{*}(i_k,j_k)-\stackrel{\‾}{w_k^{*}}))}{\sqrt{\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}{(w_k(i_k,j_k)-\stackrel{\‾}{w_k})}^2\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}{(w_k^{*}(i_k,j_k)-\stackrel{\‾}{w_k^{*}})}^2}},$ 1≤k≤K, where w _k (i _k , j _k ) represents the pixel value of the pixel whose coordinate position is (i _k , j _k ) in the original kth binary digital watermark W _k ,

Indicates the extracted kth binary digital watermark

The pixel value of the pixel whose coordinate position is (i _k , j _k ), and

represent the digital watermark W _k and

The mean of the pixel values of all pixels in . According to the size of the normalized correlation coefficient ρ, it can be judged whether to extract the embedded binary digital watermark and how robust the embedded binary digital watermark is.

Figure 2a, Figure 2b and Figure 2c show the embedded binary signature watermark W ₁ , embedded binary signature watermark W ₁ and binary serial number watermark W ₂ and embedded binary signature watermark W ₁ , binary serial number watermark W ₂ and Lena digital image after binary icon watermark W ₃ . From Figure 2a, Figure 2b and Figure 2c, it can be seen that the quality of the Lena digital image after embedding the watermark does not change (PSNR→∞dB), which is consistent with the original Lena digital image and fully meets the watermark imperceptibility requirements. At the same time, it can be seen that with the multiple zero-watermark embedding method of the present invention, the quality of the Lena digital image after embedding the watermark will not change with the number of embedded watermarks, so in practical applications, multiple binary digital images can be conveniently and flexibly embedded as required watermark. Figure 2d, Figure 2e, and Figure 2f are the three binary digital watermarks extracted from Figure 2c respectively. When the watermarked Lena digital image shown in Figure 2c is not subjected to any attack processing, the embedded three-level digital watermark can be extracted completely losslessly. A binary digital watermark, the normalized correlation coefficient ρ is 1.

In the following, a variety of attack processing is performed on the watermarked Lena digital image to verify the robustness of the multiple zero-watermark embedding methods proposed by the present invention.

(1) Brightness change

Perform brightness adjustment processing on the watermarked Lena digital image shown in Figure 2c, that is, add 0.2, 0.5 and subtract 0.2, 0.5 to all pixel values, and the corresponding results are as shown in Figure 3a, Figure 3b, Figure 3c and Figure 3d Watermark Lena digital image. After the addition and subtraction of image pixel values, visually, the brightness and darkness of the watermarked Lena digital images shown in Figure 3a, Figure 3b, Figure 3c and Figure 3d have changed significantly, while the peak signal-to-noise ratio (PSNR) has decreased respectively to 13.97dB, 6.02dB, 13.97dB and 6.02dB. Watermark extraction is performed on the watermarked Lena digital images in Figure 3a, Figure 3b, Figure 3c and Figure 3d respectively, and the extracted three binary digital watermarks are shown in Figure 3e, Figure 3f and Figure 3g respectively. The results show that the watermarks extracted from the four watermarked Lena digital images are not only identical, but also completely consistent with the original three binary digital watermarks. It can be seen that the digital watermark embedded by the method of the present invention is not affected by the brightness change of the digital image at all, and can be extracted correctly.

(2) Histogram equalization

Perform histogram equalization processing on the watermarked Lena digital image shown in Figure 2c to obtain the watermarked Lena digital image shown in Figure 4a. After histogram equalization, the pixel value distribution of the watermarked Lena digital image has changed significantly, and the peak signal-to-noise ratio (PSNR) has dropped to 19.56dB. Figure 4b, Figure 4c and Figure 4d are three binary digital watermarks extracted from the watermarked Lena digital image in Figure 4a respectively. From the results, it can be seen that the embedded three binary digital watermarks can be ideally extracted. The normalized correlation coefficient ρ reached 0.990, 0.991 and 0.997, respectively.

(3) Median filtering

Median filtering is performed on the watermarked Lena digital image shown in Figure 2c, and the filter window sizes are [5×5] and [11×11] respectively. The watermarked Lena digital image obtained after filtering is shown in Figure 5a and Figure 5b respectively. Show. Figure 5c, Figure 5d and Figure 5e are three binary digital watermarks extracted from the watermark Lena digital image in Figure 5a respectively, and Figure 5f, Figure 5g and Figure 5h are the watermarks extracted from the watermark Lena digital image in Figure 5b The three binary digital watermarks of . From the results of watermark extraction, we can see that for the median filter processing of the [5×5] small window, the embedded three binary digital watermarks are not affected in any way, and have ideal robustness, while for the [11×11] large window Median filtering process, as can be seen from Figure 5b, at this time, the details of the watermarked Lena digital image are very blurred, and the peak signal-to-noise ratio PSNR drops to 25.83dB, but the embedded three binary digital watermarks have very ideal anti-filtering The processing capacity and the normalized correlation coefficient ρ reached 0.964, 0.991 and 0.997 respectively.

Table 1 specifically shows the watermarked Lena digital image and its watermark extraction results after median filtering with different window sizes. It can be seen from the table that the multiple zero-watermark embedding methods proposed by the present invention have very ideal anti-filtering processing capabilities.

Table 1 Watermark Lena digital image quality and extraction results after median filter with different window sizes

(4) JPEG lossy compression

The watermarked Lena digital image shown in Figure 2c is subjected to JPEG lossy compression processing, and the compression quality factors are 10% and 4%, respectively, and the obtained watermarked Lena digital images are shown in Figure 6a and Figure 6b, respectively. Figure 6c, Figure 6d and Figure 6e are three binary digital watermarks extracted from the watermark Lena digital image in Figure 6a respectively, and Figure 6f, Figure 6g and Figure 6h are the watermarks extracted from the watermark Lena digital image in Figure 6b respectively The three binary digital watermarks of . From the results of watermark extraction, it can be seen that for the JPEG lossy compression process with a compression quality factor of 10%, the embedded three binary digital watermarks are not affected in any way and have ideal robustness, while for the JPEG with a compression quality factor of 4%. Lossy compression processing, as can be seen from Figure 6b, at this time the watermarked Lena digital image presents a very obvious block effect, the visual quality is seriously degraded, and the peak signal-to-noise ratio PSNR is only 25.92dB, but the embedded three binary values Digital watermarking still has ideal anti-JPEG lossy compression processing ability, and the normalized correlation coefficient ρ has reached 0.985, 0.997 and 1.0 respectively.

Table 2 specifically gives the watermarked Lena digital image quality and watermark extraction results under different JPEG compression quality factors. It can be seen from Table 2 that the method of the present invention has a very ideal anti-JPEG compression processing ability, and can still extract the three embedded binary digital watermarks without error when the compression quality factor is reduced to 6%.

Table 2 Watermark Lena digital image quality and extraction results under different JPEG compression quality factors

(5) Superimposed Gaussian noise

Noise interference is performed on the watermarked Lena digital image shown in Figure 2c. The noise is selected from two Gaussian noises with a mean value of 0 and a variance of 0.02 and a mean value of 0 and a variance of 0.05. The obtained watermarked Lena digital images are shown in Figure 7a and Figure 7b respectively . Figure 7c, Figure 7d and Figure 7e are three binary digital watermarks extracted from the watermark Lena digital image in Figure 7a respectively, and Figure 7f, Figure 7g and Figure 7h are the watermarks extracted from the watermark Lena digital image in Figure 7b The three binary digital watermarks of . It can be seen from the watermark extraction results that although the watermarked Lena digital image is interfered by Gaussian noise, the visual quality is seriously degraded, and the peak signal-to-noise ratio PSNR is only 17.22dB and 13.67dB. But for the Gaussian noise interference with the mean value of 0 and the variance of 0.02, the embedded three binary digital watermarks are not affected in any way, which has ideal robustness, while for the Gaussian noise interference with the mean value of 0 and the variance of 0.05, the embedded three Binary digital watermarking still has ideal anti-noise ability, and the normalized correlation coefficient ρ has reached 0.980, 0.994 and 1.0 respectively.

Table 3 specifically shows the watermark Lena digital image quality and watermark extraction results under the interference of Gaussian noise with a mean of 0 and different variances. It can be seen from Table 3 that the method of the present invention has a very ideal anti-noise ability, and for Gaussian noise interference with a variance less than 0.03, the three embedded binary digital watermarks can be extracted intact.

Table 3 Watermark Lena digital image quality and extraction results under different Gaussian noise intensities

(6) Geometric cutting

Geometrically cut the watermarked Lena digital image shown in Figure 2c, and cut off 128×128 and 256×256 pixels from the upper left corner, respectively, to obtain the watermarked Lena digital image shown in Figure 8a and Figure 8b. Figure 8c, Figure 8d and Figure 8e are three binary digital watermarks extracted from the watermark Lena digital image in Figure 8a respectively, and Figure 8f, Figure 8g and Figure 8h are the watermarks extracted from the watermark Lena digital image in Figure 8b The three binary digital watermarks of . It can be known from the watermark extraction results that although the watermarked Lena digital image is damaged to a large extent, and the peak signal-to-noise ratio PSNR is only 17.29dB and 11.14dB, the method of the present invention is quite robust to geometric cutting, and the embedded three The binary digital watermark can still be extracted well, and the normalized correlation coefficient ρ reaches 0.973, 0.972 and 0.981 and 0.874, 0.883 and 0.899 respectively.

(7) Geometric rotation

The watermarked Lena digital image shown in Fig. 2c is rotated counterclockwise, and the angles are 5 degrees and 25 degrees respectively. In order to extract the watermark, the rotated image is reversely rotated to restore the original direction. The watermarked Lena digital images obtained at this time are shown in Figure 9a and Figure 9b respectively, and their peak signal-to-noise ratios PSNR are 19.10dB and 13.80dB. Figure 9c, Figure 9d and Figure 9e are three binary digital watermarks extracted from the watermark Lena digital image in Figure 9a respectively, and Figure 9f, Figure 9g and Figure 9h are the watermarks extracted from the watermark Lena digital image in Figure 9b respectively The three binary digital watermarks of . It can be seen from the watermark extraction results that the method of the present invention is also relatively robust to geometric rotation attacks, and the embedded three binary digital watermarks can still be clearly extracted, and the normalized correlation coefficients ρ have reached 0.819 and 0.829 respectively and 0.871 and 0.694, 0.751 and 0.817.

Claims (4)

1. a plurality of zero watermark embedding methods of digital image, it is characterized in that comprising the following steps: ①-1. At multiple zero-watermark embedding terminals, assume that the original digital image to be embedded with K binary digital watermarks is an 8-bit grayscale image, and denoted as F, F={0≤f(m,n)≤255 ,1≤m≤M,1≤n≤N}, where K≥2, M represents the vertical resolution of the original digital image F to be embedded with K binary digital watermarks, and N represents the K binary digital watermarks to be embedded The horizontal resolution of the original digital image F of the watermark, M×N represents the resolution of the original digital image F to be embedded with K binary digital watermarks, f(m,n) represents the original digital image F to be embedded with K binary digital watermarks The pixel value of the pixel whose coordinate position is (m, n) in the image F; ①-2. At multiple zero-watermark embedding terminals, assuming that the K binary digital watermarks to be embedded are all binary images, which are recorded as W ₁ , W ₂ ,...,W _k ,...,W _K , then for The kth binary digital watermark W _k to be embedded, W _k ={w _k (i _k , j _k )=0 or 1, 1≤i _k ≤I _k ,1≤j _k ≤Jk}, where, 1 ≤k≤K, I _k represents the vertical resolution of the kth binary digital watermark W _k to be embedded, J _k represents the horizontal resolution of the kth binary digital watermark W _k to be embedded, I _k ×J _k represents the resolution of the kth binary digital watermark W _k to be embedded, w _k (i _k , j _k ) represents the coordinate position of the kth binary digital watermark W _k to be embedded is (i _k , j _k ) pixel value of the pixel; ①-3. The original digital image F to be embedded with K binary digital watermarks is normalized to obtain the normalized digital image, denoted as F′, and the normalized digital image F′ is The pixel value of the pixel whose coordinate position is (m,n) is recorded as f'(m,n), f'(m,n)=f(m,n)/255; ①-4. Carry out L-level two-dimensional discrete wavelet transform to F′ to obtain a first wavelet approximation subgraph and multiple first wavelet detail subgraphs, and denote the first wavelet approximation subgraph as FA, where the resolution of FA The rate is (M/2 ^L )×(N/2 ^L ),

min() is the minimum value function, max() is the maximum value function, the symbol

Indicates the largest integer smaller than itself, I ₁ represents the vertical resolution of the first binary digital watermark W ₁ to be embedded, J ₁ represents the horizontal resolution of the first binary digital watermark W ₁ to be embedded, I _K represents the vertical resolution of the Kth binary digital watermark W _K to be embedded, and J _K represents the horizontal resolution of the Kth binary digital watermark W _K to be embedded; ①-5. Carry out two-dimensional discrete cosine transform to FA to obtain a first two-dimensional discrete cosine transform coefficient matrix with the same resolution as FA, denoted as FAC, and then carry out Zig-Zag scanning arrangement on FAC to obtain a first one dimensional discrete cosine transform coefficient sequence, denoted as FACS, FACS={facs(x),1≤x≤(M/2 ^L )×(N/2 ^L )}, where facs(x) represents the x-th in FACS Discrete cosine transform coefficients, the second discrete cosine transform coefficients in FACS are initially all discrete cosine transform AC coefficients; ①-6. According to the resolution of each binary digital watermark to be embedded, select I ₁ ×J ₁ , I ₂ ×J ₂ ,…, I _k ×J _k , ...and I _K ×J _K discrete cosine transform AC coefficients to form K first one-dimensional discrete cosine transform AC coefficient sequences, respectively denoted as FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K , for from The kth first one-dimensional discrete cosine transform AC coefficient sequence FACS k composed of I _k ×J _k discrete cosine transform AC coefficients selected in FACS that _{meet the set conditions, FACS k =} {facs _k (y),1 ≤y≤I _k ×J _k }, and then record the corresponding position information of each discrete cosine transform AC coefficient in FACS in FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K , where I ₂ represents the The vertical resolution of the second binary digital watermark W ₂ , J ₂ represents the horizontal resolution of the second binary digital watermark W ₂ to be embedded, and facs _k (y) represents the yth discrete value in FACS _k For the cosine transform AC coefficient, the setting condition is such that the absolute value of the difference between any two adjacent discrete cosine transform AC coefficients in the formed first one-dimensional discrete cosine transform AC coefficient sequence is greater than or equal to the set difference threshold; ①-7. Compare the size of any two adjacent discrete cosine transform AC coefficients in FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K respectively, and return a logical value 1 or 0 according to the comparison result, for FACS _k Any two adjacent discrete cosine transform AC coefficients in , record them as facs _k (z) and facs _k (z+1) respectively, and judge whether facs _k (z)>facs _k (z+1) is true, if If it is true, return a logic value of 1, otherwise, return a logic value of 0, where, 1≤z≤I _k ×J _k -1; then compare the last one of FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K DCT AC coefficient and the size of the first DCT AC coefficient, and return a logic value of 1 or 0 according to the comparison result. For the last DCT AC coefficient and the first DCT AC coefficient in FACS _k , If _the former is greater, return _a logical value of 1, otherwise, return a logical value of ₀ ; then construct a one-to-one correspondence of the first and _second value digital watermark key, for the returned logical value corresponding to FACS _k , the returned logical value is stored in a two-dimensional matrix with a size of I _k × J _k in the order of first row and second column, and the two-dimensional matrix is used as The k-th first binary digital watermark key, denoted as WB _k ; ①-8. The K binary digital watermarks W ₁ , W ₂ ,...,W _k ,...,W _K to be embedded are respectively scrambled, and the K binary digital watermarks obtained after the scrambled processing are respectively corresponding to are WS ₁ , WS ₂ , ..., WS _k , ... and _WSK , then combine WS ₁ , WS ₂ , ..., WS _k , ... and WS _K with K first binary digital watermark keys WB ₁ , WB ₂ , ..., WB _k , ... and WB _K are one-to-one XORed to obtain K zero-watermark information, respectively recorded as WO ₁ , WO ₂ , ..., WO _k , ... and WO _K , WO ₁ =xor( WS ₁ ,WB ₁ ), WO ₂ =xor(WS ₂ ,WB ₂ ),…, WO _k =xor(WS _k ,WB _k ),…, WO _K =xor(WS _K ,WB _K ), and K Zero watermark information WO ₁ , WO ₂ , ..., WO _k , ... and WO _K are stored in the digital watermark information database of the registration authority, and K binary digital watermarks W ₁ , W ₂ , ..., W _k , ..., The embedding of W _K , where WS ₁ represents the binary digital watermark obtained after scrambling W ₁ , WS ₂ represents the binary digital watermark obtained after scrambling W ₂ , and WS _k represents the binary watermark obtained after scrambling W _k The binary digital watermark obtained after scrambling processing, WS _K represents the binary digital watermark obtained after scrambling W _K , WB ₁ represents the first first binary digital watermark key, WB ₂ represents the second The first binary digital watermark key, WB _K represents the Kth first binary digital watermark key, and xor() is an exclusive OR operation function; ①-9. At multiple zero-watermark embedding terminals, record the position information of each discrete cosine transform AC coefficient in FACS in FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K , and K zero-watermark information WO ₁ , WO ₂ ,..., WO _k ,..., WO _K and K binary digital watermarks W ₁ , W ₂ ,..., W _k ,..., W _K are transmitted to multiple zero-watermark extraction terminals. 2. A plurality of zero watermark embedding methods of a digital image according to claim 1, characterized in that any of FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K in the described step 1.-6 The absolute value of the difference between two adjacent discrete cosine transform AC coefficients is greater than or equal to δ ₁ , δ ₂ , ..., δ _k , ... and δ _K , where δ ₁ means that for any two adjacent discrete cosine transforms in FACS ₁ The first difference threshold set by the absolute value of the difference of the discrete cosine transform AC coefficients, δ ₂ represents the second threshold set for the absolute value of the difference between any two adjacent discrete cosine transform AC coefficients in FACS ₂ difference threshold, δ _k represents the kth difference threshold set for _the absolute value of the difference between any two adjacent discrete cosine transform AC coefficients in FACS _k , and δ _K represents the kth difference threshold for any adjacent The absolute value of the difference between the two DCT AC coefficients sets the Kth difference threshold. 3. A method for extracting multiple zero watermarks of a digital image, characterized in that it comprises the following steps: ②-1. At multiple zero-watermark extraction terminals, record the digital images of K binary digital watermarks to be extracted as TF, TF={0≤tf(m′,n′)≤255, 1≤m′≤M ', 1≤n'≤N'}, where K≥2, M' represents the vertical resolution of the digital image TF of the K binary digital watermarks to be extracted, and N' represents the resolution of the K binary digital watermarks to be extracted The horizontal resolution of the digital image TF, M'×N' represents the resolution of the digital image TF to be extracted K binary digital watermarks, the resolution of the digital image TF to be extracted K binary digital watermarks and multiple zero watermarks The resolution of the digital image embedded with the digital watermark at the embedding end is the same, and tf(m',n') represents the pixel whose coordinate position is (m',n') in the digital image TF of the K binary digital watermark to be extracted value; ②-2. At multiple zero-watermark extraction terminals, mark the K binary digital watermarks to be extracted as W′ ₁ , W′ ₂ , ..., W′ _k , ... and W′ _K . k binary digital watermarks W′ _k , W′ _k ={w′ _k (i′ _k ,j′ _k )=0 or 1, 1≤i′ _k ≤I′ _k ,1≤j′ _k ≤J′ _k }, where, 1≤k≤K, I′ _k represents the vertical resolution of the kth binary digital watermark W′ _k to be extracted, and J′ _k represents the kth binary digital watermark W′ to be extracted The horizontal resolution of _k , I′ _k ×J′ _k represents the resolution of the kth binary digital watermark W′ _k to be extracted, and the resolution of the kth binary digital watermark W′ _k to be extracted is related to multiple The resolution of the kth binary digital watermark embedded in the zero watermark embedding end is the same, w′ _k (i′ _k , j′ _k ) means that the coordinate position of the kth binary digital watermark W′ _k to be extracted is (i ′ _k , j′ _k ) pixel value of the pixel; ②-3. Perform normalization processing on the digital image TF to be extracted K binary digital watermarks, and obtain the digital image after normalization processing, which is denoted as TF′, and the coordinates in the digital image TF′ after normalization processing The pixel value of the pixel at position (m',n') is recorded as tf'(m',n'), tf'(m',n')=tf(m',n')/255; ②-4. Carry out L'-level two-dimensional discrete wavelet transform to TF' to obtain a second wavelet approximation subgraph and a plurality of second wavelet detail subgraphs, and denote the second wavelet approximation subgraph as TFA, wherein the TFA's The resolution is (M′/2 ^L ′)×(N′/2 ^L ′),

min() is the minimum value function, max() is the maximum value function, the symbol

Indicates the largest integer smaller than itself, I ₁ ′ represents the vertical resolution of the first binary digital watermark W′ ₁ to be extracted, and J ₁ ′ represents the resolution of the first binary digital watermark W′ ₁ to be extracted Horizontal resolution, I _K ′ represents the vertical resolution of the Kth binary digital watermark W′ _K to be extracted, and J _K ′ represents the horizontal resolution of the Kth binary digital watermark W′ _K to be extracted; ②-5. Perform two-dimensional discrete cosine transform on TFA to obtain a second two-dimensional discrete cosine transform coefficient matrix with the same resolution as TFA, denoted as TFAC, and then carry out Zig-Zag scanning arrangement on TFAC to obtain a second one dimensional discrete cosine transform coefficient sequence, denoted as TFACS, TFACS={tfacs(x′),1≤x′≤(M′/2 ^L ′)×(N′/2 ^L ′)}, where, tfacs(x′ ) represents the x′th discrete cosine transform coefficient in TFACS, and the second discrete cosine transform coefficient in TFACS is initially an AC discrete cosine transform coefficient; ②-6. According to the corresponding position information in FACS of each discrete cosine transform AC coefficient in FACS ₁ , FACS ₂ , ..., FACS _k , ... and FACS _K recorded by multiple zero-watermark embedding terminals, extract from TFACS respectively The I ₁ ′×J ₁ ′, I ₂ ′×J ₂ ′, ..., I _k ′×J _k ′, ... and I _K ′×J _K ′ discrete cosine transform AC coefficients at the corresponding positions constitute K second The one-dimensional discrete cosine transform AC coefficient sequence is correspondingly denoted as TFACS ₁ , TFACS ₂ , ..., TFACS _k , ... and TFACS _K , for each discrete cosine transform AC coefficient in FACS _k recorded according to multiple zero-watermark embedding terminals in Corresponding position information in FACS, extract the kth second one-dimensional discrete cosine transform AC coefficient sequence TFACS k composed of I _k ′×J _k ′ discrete cosine transform AC coefficients corresponding to the position from TFACS, TFACS _{k =} { tfacs _k (y′), 1≤y′≤I _k ′×J _k ′}, where I _k ′ represents the vertical resolution of the kth binary digital watermark W′ _k to be extracted, and J _k ′ represents The horizontal resolution of the kth binary digital watermark W′ _k to be extracted, tfacs _k (y′) represents the y′ discrete cosine transform AC coefficient in TFACS _k ; ②-7. Compare the size of any two adjacent discrete cosine transform AC coefficients in TFACS ₁ , TFACS ₂ , ..., TFACS _k , ... and TFACS _K respectively, and return a logical value 1 or 0 according to the comparison result. For TFACS _k Any two adjacent discrete cosine transform AC coefficients in , record them as tfacs _k (z′) and tfacs _k (z′+1) respectively, judge tfacs _k (z′)>tfacs _k (z′+1) Whether it is true, if true, return a logical value 1, otherwise, return a logical value 0, where, 1≤z′≤I _k ′×J _k ′-1; then compare TFACS ₁ , TFACS ₂ , ..., TFACS _k , ... and the size of the last DCT AC coefficient and the first DCT coefficient in TFACS _K , and return a logical value 1 or 0 according to the comparison result, for the last DCT AC coefficient and the first DCT coefficient in TFACS _k A discrete cosine transform AC coefficient, if the _{former is} large, return a logical value of 1, otherwise, return _a logical value _of 0; then construct One-to-one correspondence with the second binary digital watermark key, for the returned logical value corresponding to TFACS _k , store the returned logical value in a two-dimensional array with a size of I _k ′×J _k ′ In the matrix, the two-dimensional matrix is used as the kth second binary digital watermark key, denoted as TWB _k ; ②-8. Combine K pieces of zero-watermark information WO ₁ , WO ₂ , ..., WO _k , ... and WO _K from multiple zero-watermark embedding terminals with K second binary digital watermark keys TWB ₁ , TWB ₂ respectively , ..., TWB _k , ... and TWB _K are one-to-one XOR operation, and K binary digital watermarks are recovered, which are respectively recorded as TW ₁ , TW ₂ , ..., TW _k , ... and TW _K , TW ₁ = xor(WO ₁ ,TWB ₁ ), TW ₂ =xor(WO ₂ ,TWB ₂ ),…,TW _k =xor(WO _k ,TWB _k ),…,TW _K =xor(WO _K ,TWB _K ), where , TWB ₁ represents the first second binary digital watermark key, TWB ₂ represents the second second binary digital watermark key, TWB _k represents the k-th second binary digital watermark key, TWB _K represents the second K second binary digital watermark keys, xor () is an exclusive OR operation function; ②-9. Perform anti-scrambling processing on K binary digital watermarks TW ₁ , TW ₂ , ..., TW _k , ... and TW _K respectively, to obtain K binary digital watermarks with copyright authentication information, which are respectively recorded as

…and

②-10. K binary digital watermarks with copyright authentication information

…and

Carry out similarity calculation in one-to-one correspondence with K binary digital watermarks W ₁ , W ₂ , ..., W _k , ... and W _K embedded in multiple zero-watermark embedding terminals, and obtain K normalized correlation coefficients correspondingly, and then Determine whether to extract K binary digital watermarks W ₁ , W ₂ , . . . , W _{k ,} . 4. a plurality of zero watermark extracting methods of a kind of digital image according to claim 3 is characterized in that the concrete process of described step 2.-10 is: z1. K binary digital watermarks with copyright authentication information

…and

Carry out one-to-one similarity calculation with the K binary digital watermarks W ₁ , W ₂ , ..., W _k , ... and W _K embedded in multiple zero-watermark embedding terminals, and obtain K normalized correlation coefficients correspondingly. For Will The kth normalized correlation coefficient obtained after similarity calculation with Wk is recorded as $\mathrm{\ρ}(W_k,W_k^{*})=\frac{\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}((w_k(i_k,j_k)-\stackrel{\‾}{w_k})\×(w_k^{*}(i_k,j_k)-\stackrel{\‾}{w_k^{*}}))}{\sqrt{\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}{(w_k(i_k,j_k)-\stackrel{\‾}{w_k})}^2\×\underset{i_k=1}{\overset{I_k}{\mathrm{\Σ}}}\underset{j_k=1}{\overset{J_k}{\mathrm{\Σ}}}{(w_k^{*}(i_k,j_k)-\stackrel{\‾}{w_k^{*}})}^2}},$ Among them, w _k (i _k , j _k ) represents the pixel value of the pixel whose coordinate position is (i _k , j _k ) in the kth binary digital watermark W _k embedded by multiple zero-watermark embedding terminals,

Represents the mean value of the pixel values of all pixels in the kth binary digital watermark W _k embedded in multiple zero-watermark embedding terminals,

Represents the kth binary digital watermark with copyright authentication information

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

Represents the kth binary digital watermark with copyright authentication information

The mean of the pixel values of all pixels in ; z2. Determine whether to extract K binary digital watermarks W ₁ , W ₂ , ..., W _k , ... and W _K embedded in multiple zero-watermark embedding terminals according to the size of the K normalized correlation coefficients. correlation coefficient

if

A value of 1 determines that W _k is extracted losslessly, if

The value of is greater than or equal to δ _T and less than 1, then it is determined that the extraction of W _k is successful, if

If the value of is less than δ _T , it is determined that W _k extraction fails, where δ _T represents the watermark extraction threshold.

Images