CN102223561A - Blind watermark embedding and extracting method of stereoscopic video image - Google Patents

Abstract

The invention discloses a method for blind watermark embedding and extraction of a stereoscopic video image. By dividing the stereoscopic video image into non-overlapping image blocks, performing discrete cosine transform and secondary discrete wavelet transform on each image block, and analyzing the stereoscopic video image The size relationship between the discrete cosine transform DC coefficients of the left and right viewpoint images and the low-frequency coefficients of the second-level discrete wavelet transform, so as to define the internal relationship of the image block and the inter-block relationship between the two image blocks in the stereoscopic video image, through these Whether the relationship and the watermark information are consistent determines the watermark embedding method, wherein the data of the original stereoscopic video image is not changed when the relationship is embedded, and the DC coefficient of the image block whose DC coefficient is modified by the discrete cosine transform is modified when the quantization is embedded, so that The quality of the stereoscopic video image is effectively guaranteed; at the same time, the original stereoscopic video image is not needed when the watermark is extracted, and blind detection is realized.

Description

A Blind Watermark Embedding and Extraction Method for Stereo Video Images

technical field

The invention relates to a digital watermark technology, in particular to a method for blind watermark embedding and extraction of stereoscopic video images.

Background technique

Stereoscopic video images show the information of a certain scene or object at different angles at the same moment, which increases the depth information of the scene and enhances the sense of visual reality and lifelikeness. Therefore, the stereoscopic video system has broad application prospects in fields such as stereoscopic television, three-dimensional video conferencing, telemedicine and special effect advertisements. With the advent of three-dimensional movies, the illegal duplication and dissemination of some high-end works have threatened intellectual property rights. Therefore, copyright protection measures for stereoscopic video images are very necessary.

Digital watermarking technology can well embed certain secret information in digital products such as digital images, audio and video, so as to protect the copyright of digital products, prove the authenticity of digital products, track piracy or provide additional information of products. According to different applications, digital watermarking can be divided into robust watermarking, fragile watermarking and semi-fragile watermarking. Robust watermarks are used for copyright protection, while fragile watermarks and semi-fragile watermarks are used for integrity verification of video or multimedia data. However, the existing digital watermarking technology is mainly aimed at two-dimensional images, and how to apply digital watermarking technology to stereoscopic video images has become a problem that needs to be broken through.

Contents of the invention

The technical problem to be solved by the present invention is to provide a blind watermark embedding and extraction method that can effectively protect image copyright on the premise that the quality of stereoscopic video images remains basically unchanged or does not change much.

The technical solution adopted by the present invention to solve the above-mentioned technical problems is: a blind watermark embedding method for stereoscopic video images, which is characterized in that it comprises the following steps:

将原始立体视频图像S_org的右视点图像R_org中的第i个图像块记为

原始立体视频图像S_org的左视点图像L_org中的第i个图像块

和原始立体视频图像S_org的右视点图像R_org中的第i个图像块

构成原始立体视频图像S_org中的第i个图像块对，记为

其中，M表示原始立体视频图像S_org的宽，N表示原始立体视频图像S_org的高，1≤i≤((M×N)/(n×n))；①At the watermark embedding end, let S _org be the original stereoscopic video image with size M×N, the original stereoscopic video image S _org includes left viewpoint image and right viewpoint image, denote the left viewpoint image of the original stereoscopic video image S _org as L _org , record the right viewpoint image of the original stereoscopic video image S _org as R _org , divide the left viewpoint image L _org of the original stereoscopic video image S _org and the right viewpoint image R _org of the original stereoscopic video image S _org into (M×N)/ (n×n) non-overlapping image blocks of size n×n, the i-th image block in the left view image L _org of the original stereoscopic video image S _org is recorded as

Denote the i-th image block in the right view image R _org of the original stereoscopic video image S _org as

The i-th image block in the left view image L _org of the original stereoscopic video image S _org

and the i-th image block in the right view image R _org of the original stereoscopic video image S _org

Constitute the i-th image block pair in the original stereoscopic video image S _org , denoted as

Wherein, M represents the width of the original stereoscopic video image S _org , N represents the height of the original stereoscopic video image S _org , 1≤i≤((M×N)/(n×n));

其中，m×m≤((M×N)/(n×n))；②Let W be a binary watermark image with a size of m×m, perform scrambling transformation on the binary watermark image W, and obtain the scrambled binary watermark image, denoted as

Among them, m×m≤((M×N)/(n×n));

3. Add watermarks to each image block in the original stereoscopic video image S _org in turn, and its specific process is as follows:

为当前图像块对；③-1. Define the i-th image block pair in the original stereoscopic video image S _org currently being processed

is the current image block pair;

分别进行离散余弦变换和二级离散小波变换，对当前图像块对

中的分别进行离散余弦变换和二级离散小波变换；③-2. For the current image block pair middle

Discrete cosine transform and two-level discrete wavelet transform are performed respectively, and the current image block pair

middle Carry out discrete cosine transform and secondary discrete wavelet transform respectively;

的内部关系，判断

的离散余弦变换直流系数是否大于

的二级离散小波变换的设定位置的低频系数，如果是，则置

的内部关系Intra-L为“1”，否则，置

的内部关系Intra-L为“0”；令Intra-R表示

的内部关系，判断

的离散余弦变换直流系数是否大于

的二级离散小波变换的设定位置的低频系数，如果是，则置

的内部关系Intra-R为“1”，否则，置

的内部关系Intra-R为“0”；令Intra表示

的内部关系，判断

的内部关系Intra-L和

的内部关系Intra-R是否均为“1”，如果是，则置

的内部关系Intra为“1”，否则，置

的内部关系Intra为“0”；令Inter表示的块间关系，判断

的离散余弦变换直流系数是否大于

的离散余弦变换直流系数，如果是，则置

的块间关系Inter为“1”，否则，置

的块间关系Inter为“0”；③-3. Let Intra-L represent

internal relationship, judgment

Is the discrete cosine transform DC coefficient greater than

The low-frequency coefficient of the set position of the second-level discrete wavelet transform, if it is, then set

The internal relationship Intra-L is "1", otherwise, set

The internal relationship Intra-L of is "0"; let Intra-R represent

internal relationship, judgment

Is the discrete cosine transform DC coefficient greater than

The low-frequency coefficient of the set position of the second-level discrete wavelet transform, if it is, then set

The internal relationship Intra-R is "1", otherwise, set

The internal relationship Intra-R of is "0"; let Intra represent

internal relationship, judgment

The internal relationship of Intra-L and

Whether the internal relations Intra-R of are all "1", if yes, then set

The internal relation of Intra is "1", otherwise, set

The internal relationship Intra of is "0"; let Inter represent The inter-block relationship, judging

Is the discrete cosine transform DC coefficient greater than

DC coefficient of discrete cosine transform, if yes, then set

Inter-block relationship Inter is "1", otherwise, set

The inter-block relationship Inter is "0";

的二级离散小波变换的设定位置的低频系数与

的二级离散小波变换的设定位置的低频系数为各自图像块的二级离散小波变换的相同位置的低频系数；the above

The low-frequency coefficients of the set position of the second-order discrete wavelet transform with

The low-frequency coefficients at the set position of the second-level discrete wavelet transform are the low-frequency coefficients at the same position of the second-level discrete wavelet transform of the respective image blocks;

的内部关系Intra的值与置乱后的二值水印图像

中的第i个二值像素点的像素值w_i是否相同，如果相同，则不修改

和

的离散余弦变换直流系数，并标记当前图像块对

的水印嵌入方式为00，然后执行步骤③-7；否则，执行步骤③-5；③-4. Judging the current image block pair

The value of the internal relationship Intra and the binary watermarked image after scrambling

Whether the pixel value w _i of the i-th binary pixel point in is the same, if it is the same, it will not be modified

and

The DC coefficients of the discrete cosine transform, and mark the current image block pair

The embedding mode of the watermark is 00, then perform step ③-7; otherwise, perform step ③-5;

的块间关系Inter的值与置乱后的二值水印图像

中的第i个二值像素点的像素值w_i是否相同，如果相同，则不修改

和的离散余弦变换直流系数，并标记当前图像块对

的水印嵌入方式为01，然后执行步骤③-7；否则，执行步骤③-6；③-5. Judging the current image block pair

The value of inter-block relationship Inter and the binary watermark image after scrambling

Whether the pixel value w _i of the i-th binary pixel point in is the same, if it is the same, it will not be modified

and The DC coefficients of the discrete cosine transform, and mark the current image block pair

The embedding method of the watermark is 01, and then perform step ③-7; otherwise, perform step ③-6;

的离散余弦变换直流系数按照量化步长S分别进行量化，分别得到量化值Q_L，i和Q_R，i， $Q_{L,i}=\mathrm{floor}\left(\frac{{\mathrm{DC}}_{L,i}}{S}\right),$ $Q_{R,i}=\mathrm{floor}\left(\frac{{\mathrm{DC}}_{R,i}}{S}\right),$ 其中，DC_L，i表示

的离散余弦变换直流系数，DC_R，i表示

的离散余弦变换直流系数，floor(x)为计算得到小于等于x的最大整数函数；定义

的离散余弦变换直流系数的奇偶性为其量化值Q_L，i的奇偶性，定义

的离散余弦变换直流系数的奇偶性为其量化值Q_R，i的奇偶性，判断量化值Q_L，i和量化值Q_R，i是否同为偶数或同为奇数，如果是，则认为

和

的离散余弦变换直流系数的奇偶性一致，否则，认为和

的离散余弦变换直流系数的奇偶性不一致；③-6, right and

The discrete cosine transform DC coefficients are respectively quantized according to the quantization step size S, and the quantized values Q _{L, i} and Q _{R, i} are respectively obtained, $Q_{L,i}=\mathrm{floor}\left(\frac{{\mathrm{DC}}_{L,i}}{S}\right),$ $Q_{R,i}=\mathrm{floor}\left(\frac{{\mathrm{DC}}_{R,i}}{S}\right),$ Among them, DC _{L, i} represent

The discrete cosine transform DC coefficient, DC _{R, i} represents

DC coefficient of discrete cosine transform of , floor(x) is the largest integer function less than or equal to x calculated; definition

The parity of the discrete cosine transform DC coefficient is its quantization value Q _{L, the parity of i} , defined

The parity of the discrete cosine transform DC coefficient is the parity of the quantized value Q _{R, i} , and judges whether the quantized value Q _{L, i} and the quantized value Q _{R, i} are both even or odd, and if so, it is considered

and

The parity of DC coefficients of discrete cosine transform is consistent, otherwise, consider and

The parity of the discrete cosine transform DC coefficients is inconsistent;

中的第i个二值像素点的像素值w_i是否为“1”，如果是，则修改当前图像块对

中的任一个图像块的离散余弦变换直流系数，使修改后的

和

的离散余弦变换直流系数的奇偶性一致，并标记当前图像块对

的水印嵌入方式为10，否则，修改当前图像块对

中的任一个图像块的离散余弦变换直流系数，使修改后的

和

的离散余弦变换直流系数的奇偶性不一致，并标记当前图像块对

的水印嵌入方式为10；Judging the binary watermarked image after scrambling

Whether the pixel value w _i of the i-th binary pixel in is "1", if yes, modify the current image block pair

The discrete cosine transform DC coefficients of any image block in , so that the modified

and

The parity of the DC coefficients of the discrete cosine transform is consistent, and the current image block pair is marked

The watermark embedding mode of is 10, otherwise, modify the current image block pair

The discrete cosine transform DC coefficients of any image block in , so that the modified

and

The parity of the DC coefficients of the discrete cosine transform is inconsistent, and the current image block pair is marked

The watermark embedding mode of is 10;

③-7. Determine whether i≤m×m is true, if it is true, add 1 to the value of i, use the next image block pair in the original stereoscopic video image S _org as the current image block pair, and repeat step ③-2 Go to step ③-7, until the m×m image block pairs are processed and the watermark embedding process is ended;

④. The size of the binary watermark image, the watermark embedding method and the quantization step S are used as the key, and the watermark embedding end transmits the key to the watermark extraction end.

The low-frequency coefficient at the position set in step ③-3 is the low-frequency coefficient at any position of the secondary discrete wavelet transform of the image block.

中的任一个图像块的离散余弦变换直流系数时，选择

和

中离散余弦变换直流系数修改幅度小的那个图像块的离散余弦变换直流系数进行修改。In the described step ③-6, modify the current image block pair

When the discrete cosine transform DC coefficient of any image block in , choose

and

The discrete cosine transform direct current coefficient of the image block whose modification range is small among the discrete cosine transform direct current coefficients is modified.

A method for extracting a blind watermark using the above-mentioned blind watermark embedding method to embed a watermarked stereoscopic video image, characterized in that it comprises the following steps:

待检测的立体视频图像S_wat的左视点图像L_wat中的第j个图像块

和待检测的立体视频图像S_wat的右视点图像R_wat中的第j个图像块

构成待检测的立体视频图像S_wat中的第j个图像块对，记为

其中，M表示待检测的立体视频图像S_wat的宽，N表示待检测的立体视频图像S_wat的高，1≤j≤((M×N)/(n×n))；① At the watermark extraction end, let S _wat be the stereoscopic video image to be detected _whose size is M×N. The stereoscopic video image S _wat to be detected includes left viewpoint image and right viewpoint image. The left viewpoint image is L _wat , and the right viewpoint image of the stereoscopic video image S _wat to be detected is R _wat , and the left viewpoint image L _wat of the stereoscopic video image S _wat to be detected and the right side of the stereoscopic video image S _wat to be detected are The viewpoint image R _wat is divided into (M×N)/(n×n) non-overlapping image blocks with a size _of n×n, and the j- _th image blocks are denoted as Denote the jth image block in the right viewpoint image R _wat of the stereoscopic video image S _wat to be detected as

The jth image block in the left viewpoint image L _wat of the stereoscopic video image S _wat to be detected

and the jth image block in the right view image R _wat of the stereoscopic video image S _wat to be detected

Constitute the jth image block pair in the stereoscopic video image S _wat to be detected, denoted as

Wherein, M represents the width of the stereoscopic video image S _wat to be detected, and N represents the height of the stereoscopic video image S _wat to be detected, 1≤j≤((M×N)/(n×n));

②The watermark is extracted from each image block pair in the stereoscopic video image S _wat to be detected in turn, and the specific process is as follows:

为当前图像块对；②-1, define the jth image block pair in the stereoscopic video image S _wat to be detected currently being processed

is the current image block pair;

中的

分别进行离散余弦变换和二级离散小波变换，对当前图像块对

中的

分别进行离散余弦变换和二级离散小波变换；②-2. For the current image block pair

middle

Discrete cosine transform and two-level discrete wavelet transform are performed respectively, and the current image block pair

middle

Carry out discrete cosine transform and secondary discrete wavelet transform respectively;

的内部关系，判断

的离散余弦变换直流系数是否大于

的二级离散小波变换的设定位置的低频系数，如果是，则置

的内部关系Intra-L′为“1”，否则，置

的内部关系Intra-L′为“0”；令Intra-R′表示

的内部关系，判断

的离散余弦变换直流系数是否大于

的二级离散小波变换的设定位置的低频系数，如果是，则置的内部关系Intra-R′为“1”，否则，置

的内部关系Intra-R′为“0”；令Intra′表示

的内部关系，判断

的内部关系Intra-L′和

的内部关系Intra-R′是否均为“1”，如果是，则置

的内部关系Intra′为“1”，否则，置

的内部关系Intra′为“0”；令Inter′表示

的块间关系，判断

的离散余弦变换直流系数是否大于

的离散余弦变换直流系数，如果是，则置

的块间关系Inter′为“1”，否则，置

的块间关系Inter′为“0”；②-3. Let Intra-L′ express

internal relationship, judgment

Is the discrete cosine transform DC coefficient greater than

The low-frequency coefficient of the set position of the second-level discrete wavelet transform, if it is, then set

The internal relationship Intra-L′ of is "1", otherwise, set

The internal relationship Intra-L′ of is “0”; let Intra-R′ represent

internal relationship, judgment

Is the discrete cosine transform DC coefficient greater than

The low-frequency coefficient of the set position of the second-level discrete wavelet transform, if it is, then set The internal relation Intra-R′ of is "1", otherwise, set

The internal relation Intra-R′ of is “0”; let Intra′ represent

internal relationship, judgment

The internal relations Intra-L′ and

Whether the internal relations Intra-R′ of are all "1", if yes, then set

The internal relation Intra' of is "1", otherwise, set

The internal relationship Intra' is "0"; let Inter' represent

The inter-block relationship, judging

Is the discrete cosine transform DC coefficient greater than

DC coefficient of discrete cosine transform, if yes, then set

The inter-block relationship Inter' is "1", otherwise, set

The inter-block relationship Inter' is "0";

的二级离散小波变换的设定位置的低频系数与

的二级离散小波变换的设定位置的低频系数为各自图像块的二级离散小波变换的相同位置的低频系数；the above

The low-frequency coefficients of the set position of the second-order discrete wavelet transform with

The low-frequency coefficients at the set position of the second-level discrete wavelet transform are the low-frequency coefficients at the same position of the second-level discrete wavelet transform of the respective image blocks;

和

的离散余弦变换直流系数按照量化步长S分别进行量化，分别得到量化值Q′_L，j和Q′_R，j， $Q_{L,j}^{\′}=\mathrm{floor}\left(\frac{{{\mathrm{DC}}^{\′}}_{L,j}}{S}\right),$ $Q_{R,j}^{\′}=\mathrm{floor}\left(\frac{{{\mathrm{DC}}^{\′}}_{R,j}}{S}\right),$ 其中，DC′_L，j表示

的离散余弦变换直流系数，DC′_R，j表示

的离散余弦变换直流系数，S为量化步长，floor(x)为计算得到小于等于x的最大整数函数；②-4, right

and

The discrete cosine transform DC coefficients are respectively quantized according to the quantization step size S, and the quantized values Q′ _{L, j} and Q′ _{R, j} are respectively obtained, $Q_{L,j}^{\′}=\mathrm{floor}\left(\frac{{{\mathrm{DC}}^{\′}}_{L,j}}{S}\right),$ $Q_{R,j}^{\′}=\mathrm{floor}\left(\frac{{{\mathrm{DC}}^{\′}}_{R,j}}{S}\right),$ Among them, DC′ _{L, j} represents

The DC coefficient of discrete cosine transform, DC′ _{R, j} represents

The DC coefficient of discrete cosine transform, S is the quantization step size, and floor (x) is the largest integer function less than or equal to x calculated for;

的水印嵌入方式是否为00，如果不是，则执行步骤②-6，否则，再判断

的内部关系Intra-L′和

的内部关系Intra-R′是否均为“1”，如果均为“1”，则从当前图像块对

中提取出水印“1”，如果均为“0”或其中一个为“0”，则从当前图像块对中提取出水印“0”，然后执行步骤②-8；②-5. Judging the current image block pair

Whether the embedding method of the watermark is 00, if not, go to step ②-6, otherwise, judge again

The internal relations Intra-L′ and

Whether the internal relationship Intra-R′ of is all "1", if they are all "1", then from the current image block to

Extract the watermark "1" from the current image block, if they are all "0" or one of them is "0", then the Extract the watermark "0" from , and then perform steps ②-8;

的水印嵌入方式是否为01，如果不是，则执行步骤②-7，否则，再判断当前图像块对的块间关系Inter′是否为“1”，如果为“1”，则从当前图像块对

中提取出水印“1”，如果为“0”，则从当前图像块对

中提取出水印“0”，然后执行步骤②-8；②-6. Judging the current image block pair

Whether the embedding method of the watermark is 01, if not, execute steps ②-7, otherwise, judge the current image block pair Is the inter-block relationship Inter' of "1", if it is "1", then from the current image block to

Extract the watermark "1" from the watermark, if it is "0", then from the current image block pair

Extract the watermark "0" from , and then perform steps ②-8;

中提取出水印“1”，否则，从当前图像块对

中提取出水印“0”，然后执行步骤②-8；②-7, judge whether Q′ _{L, j} and Q′ _{R, j} are both odd numbers or even numbers, if yes, then from the current image block to

Extract the watermark "1" from the current image block, otherwise, from the current image block pair

Extract the watermark "0" from , and then perform steps ②-8;

②-8. Determine whether j≤m×m is true. If it is true, add 1 to the j value, use the next image block pair in the stereoscopic video image S _wat to be detected as the current image block pair, and repeat step ② -2 to step ②-8, until the m×m image block pairs are processed to end the watermark extraction process;

③ Perform inverse scrambling transformation on the extracted watermark image to obtain the final extracted binary watermark image.

The low-frequency coefficient at the position set in step ②-3 is the low-frequency coefficient at any position of the second-level discrete wavelet transform of the image block.

Compared with the prior art, the present invention has the advantages of dividing the stereoscopic video image into non-overlapping image blocks, implementing discrete cosine transform and secondary discrete wavelet transform on each image block, and analyzing the left and right viewpoint images of the stereoscopic video image The size relationship between the DC coefficient of the discrete cosine transform and the low-frequency coefficient of the second-level discrete wavelet transform, so as to define the internal relationship of the image block and the inter-block relationship between the two image blocks in the stereoscopic video image, through these relationships and whether the watermark information Unanimously determine the watermark embedding method, wherein the data of the original stereoscopic video image is not changed when the relationship is embedded, and the DC coefficient of the discrete cosine transform DC coefficient of the image block with a small modification range is modified when the quantization is embedded, which effectively guarantees the stereoscopic video. Image quality; at the same time, the original stereoscopic video image is not needed during watermark extraction, and blind detection is realized. When the relationship is embedded, the relationship between the DC coefficient of the discrete cosine transform in the block and the low-frequency coefficient of the second-level discrete wavelet transform and the relationship between the DC coefficients of the discrete cosine transform between the blocks are selected. The low-frequency coefficients have the same variation trend, and the relationship between the discrete cosine transform DC coefficients between the blocks has the same variation relationship under the same attack, so the stability of the relationship is strong, thereby effectively improving the robustness of the embedding method of the present invention . During relational embedding, the data of the original stereo image is not modified, thus effectively improving the quality of the watermarked stereo image.

Description of drawings

Fig. 1 a is the left viewpoint image of the first moment of " puppy " original stereoscopic video image;

Fig. 1b is the right viewpoint image of the first moment of "puppy" original stereoscopic video image;

Figure 2a is the original binary watermarked image;

Fig. 2b is the binary watermark image after the original binary watermark image shown in Fig. 2a is scrambled;

Fig. 3 a is the left viewpoint image of the first moment of the "puppy" stereoscopic video image embedded with the watermark;

Fig. 3b is the right viewpoint image at the first moment of the "puppy" stereoscopic video image embedded with the watermark;

Figure 4a is all watermarks extracted from the stereoscopic video image formed by Figure 3a and Figure 3b when there is no attack;

Fig. 4b is the binary watermark image extracted from the stereoscopic video image composed of Fig. 3a and Fig. 3b when there is no attack;

Figure 5a is a binary watermark image extracted from a JPEG-compressed stereoscopic video image with a quality factor of 90;

Figure 5b is a binary watermark image extracted from a JPEG-compressed stereoscopic video image with a quality factor of 70;

Figure 5c is a binary watermark image extracted from a JPEG-compressed stereoscopic video image with a quality factor of 50;

Figure 5d is a binary watermark image extracted from a JPEG-compressed stereoscopic video image with a quality factor of 30;

Figure 5e is a binary watermark image extracted from a JPEG-compressed stereoscopic video image with a quality factor of 20;

Figure 6a is a binary watermark image extracted from a stereoscopic video image after Gaussian filtering (window 3×3);

Fig. 6b is the binary watermark image extracted from the stereoscopic video image after median filtering (window 3×3);

Fig. 6c is the binary watermark image extracted from the stereoscopic video image after median filtering (window 5×5);

Figure 6d is a binary watermark image extracted from the stereoscopic video image after mean filtering (window 3×3);

Fig. 6e is the binary watermark image extracted from the stereoscopic video image after mean filtering (window 5×5);

Fig. 7a is the binary watermark image extracted from the stereoscopic video image cut by 1/4 from the top left end;

Figure 7b is a binary watermark image extracted from the stereoscopic video image cut by 1/8 from the top left end;

Figure 7c is a binary watermark image extracted from the stereoscopic video image cut by 1/4 in the middle;

Figure 7d is a binary watermark image extracted from the stereoscopic video image cut by 1/8 in the middle;

Figure 7e is a binary watermark image extracted from the stereoscopic video image cut by 1/2 from the top left end;

FIG. 8 is a schematic diagram of the process of sequentially adding watermarks to each image block pair in the original stereoscopic video image S _org .

Detailed ways

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

Embodiment one:

The blind watermark embedding method of a kind of stereoscopic video image that present embodiment proposes, it mainly comprises the following steps:

将原始立体视频图像S_org的右视点图像R_org中的第i个图像块记为

原始立体视频图像S_org的左视点图像L_org中的第i个图像块

和原始立体视频图像S_org的右视点图像R_org中的第i个图像块

(即左视点图像L_org和右视点图像R_org中同一坐标位置的两个图像块)构成原始立体视频图像S_org中的第i个图像块对，记为

其中，M表示原始立体视频图像S_org的宽，N表示原始立体视频图像S_org的高，1≤i≤((M×N)/(n×n))。①At the watermark embedding end, let S _org be the original stereoscopic video image with size M×N, the original stereoscopic video image S _org includes left viewpoint image and right viewpoint image, denote the left viewpoint image of the original stereoscopic video image S _org as L _org , record the right viewpoint image of the original stereoscopic video image S _org as R _org , divide the left viewpoint image L _org of the original stereoscopic video image S _org and the right viewpoint image R _org of the original stereoscopic video image S _org into (M×N)/ (n×n) non-overlapping image blocks of size n×n, the i-th image block in the left view image L _org of the original stereoscopic video image S _org is recorded as

Denote the i-th image block in the right view image R _org of the original stereoscopic video image S _org as

The i-th image block in the left view image L _org of the original stereoscopic video image S _org

and the i-th image block in the right view image R _org of the original stereoscopic video image S _org

(that is, two image blocks at the same coordinate position in the left viewpoint image L _org and the right viewpoint image R _org ) constitute the i-th image block pair in the original stereoscopic video image S _org , denoted as

Wherein, M represents the width of the original stereoscopic video image S _org , N represents the height of the original stereoscopic video image S _org , 1≤i≤((M×N)/(n×n)).

In this embodiment, the original stereoscopic video image of the test is the stereoscopic video image at the first moment in the "puppy" stereoscopic video sequence, and the resolution size is 640*480, that is, M=640, N=480, Fig. 1a and Figure 1b shows the left and right viewpoint images of the "puppy" original stereoscopic video image at the first moment, respectively.

In this embodiment, the size n×n of the image block can be selected according to the size of the stereoscopic video image, for example, 8×8, 16×16, etc. can be selected. In the actual application process, if the selection is too small, it may cause poor concealment of the embedded watermark, and if the selection is too large, the embedded watermark information will be too little. In this embodiment, n=8 is chosen.

②Let W be a binary watermark image with a size of m×m, perform scrambling transformation on the binary watermark image W, and obtain the scrambled binary watermark image, denoted as Among them, m×m≦((M×N)/(n×n)). In this embodiment, the scrambling transformation uses the Arnold scrambling transformation. The Arnold scrambling transformation is a periodic transformation, and the transformed data array must be a square matrix. The transformation period is related to the size of the square matrix, for example, 64×64 The transformation cycle of the Arnold scrambling transformation is 48, so the number of transformations L should be less than 48 to make the data achieve the purpose of scrambling. In this embodiment, the original binary watermark image with a size of 64×64 as shown in FIG. 2a is used, and the scrambled binary watermark image is shown in FIG. 2b.

3. Add watermarks to each image block in the original stereoscopic video image S _org in turn, and its specific process (Fig. 8 provides the specific watermarking process) is as follows:

为当前图像块对。③-1. Define the i-th image block pair in the original stereoscopic video image S _org currently being processed

is the current image block pair.

中的分别进行离散余弦变换和二级离散小波变换。③-2. For the current image block pair middle Carry out discrete cosine transform (Discrete Cosine Transform, DCT) and secondary discrete wavelet transform (Discrete Wavelet Transform, DWT) respectively, for the current image block pair

middle Discrete cosine transform and two-stage discrete wavelet transform are carried out respectively.

的内部关系，判断

的离散余弦变换直流系数是否大于

的二级离散小波变换的设定位置的低频系数，如果是，则置

的内部关系Intra-L为“1”，否则，置

的内部关系Intra-L为“0”；令Intra-R表示

的内部关系，判断

的离散余弦变换直流系数是否大于

的二级离散小波变换的设定位置的低频系数，如果是，则置

的内部关系Intra-R为“1”，否则，置

的内部关系Intra-R为“0”；令Intra表示

的内部关系，判断

的内部关系Intra-L和的内部关系Intra-R是否均为“1”，如果是，则置

的内部关系Intra为“1”，否则，置

的内部关系Intra为“0”；令Inter表示

的块间关系，判断

的离散余弦变换直流系数是否大于

的离散余弦变换直流系数，如果是，则置

的块间关系Inter为“1”，否则，置

的块间关系Inter为“0”。③-3. Let Intra-L represent

internal relationship, judgment

Is the discrete cosine transform DC coefficient greater than

The low-frequency coefficient of the set position of the second-level discrete wavelet transform, if it is, then set

The internal relationship Intra-L is "1", otherwise, set

The internal relationship Intra-L of is "0"; let Intra-R represent

internal relationship, judgment

Is the discrete cosine transform DC coefficient greater than

The low-frequency coefficient of the set position of the second-level discrete wavelet transform, if it is, then set

The internal relationship Intra-R is "1", otherwise, set

The internal relationship Intra-R of is "0"; let Intra represent

internal relationship, judgment

The internal relationship of Intra-L and Whether the internal relations Intra-R of are all "1", if yes, then set

The internal relation of Intra is "1", otherwise, set

The internal relationship Intra of is "0"; let Inter represent

The inter-block relationship, judging

Is the discrete cosine transform DC coefficient greater than

DC coefficient of discrete cosine transform, if yes, then set

Inter-block relationship Inter is "1", otherwise, set

Inter-block relation Inter is "0".

的二级离散小波变换的设定位置的低频系数与

的二级离散小波变换的设定位置的低频系数为各自图像块的二级离散小波变换的相同位置的低频系数。在本实施例中，上述设定位置的低频系数为图像块的二级离散小波变换的任一位置的低频系数，如此处

的二级离散小波变换的低频系数和

的二级离散小波变换的低频系数选取的是各自图像块的低频子带的第0行第1列的低频系数。the above

The low-frequency coefficients of the set position of the second-order discrete wavelet transform with

The low-frequency coefficients at the set position of the second-level discrete wavelet transform are the low-frequency coefficients at the same position of the second-level discrete wavelet transform of the respective image blocks. In this embodiment, the low-frequency coefficient at the above-mentioned setting position is the low-frequency coefficient at any position of the second-level discrete wavelet transform of the image block, as shown here

The low-frequency coefficients of the second-order discrete wavelet transform and

The low-frequency coefficients of the second-level discrete wavelet transform are selected from the low-frequency coefficients in row 0 and column 1 of the low-frequency sub-bands of the respective image blocks.

的内部关系Intra的值与置乱后的二值水印图像

中的第i个二值像素点的像素值w_i是否相同，如果相同，则不修改

和

的离散余弦变换直流系数，并标记当前图像块对

的水印嵌入方式为00，然后执行步骤③-7；否则，执行步骤③-5。③-4. Judging the current image block pair

The value of the internal relationship Intra and the binary watermarked image after scrambling

Whether the pixel value w _i of the i-th binary pixel point in is the same, if it is the same, it will not be modified

and

The DC coefficients of the discrete cosine transform, and mark the current image block pair

The embedding mode of the watermark is 00, then go to step ③-7; otherwise, go to step ③-5.

Here, the watermark embedding of this step is called relational embedding.

的块间关系Inter的值与置乱后的二值水印图像

中的第i个二值像素点的像素值w_i是否相同，如果相同，则不修改

和的离散余弦变换直流系数，并标记当前图像块对

的水印嵌入方式为01，然后执行步骤③-7；否则，执行步骤③-6。③-5. Judging the current image block pair

The value of inter-block relationship Inter and the binary watermark image after scrambling

Whether the pixel value w _i of the i-th binary pixel point in is the same, if it is the same, it will not be modified

and The DC coefficients of the discrete cosine transform, and mark the current image block pair

The embedding mode of the watermark is 01, then go to step ③-7; otherwise, go to step ③-6.

Here, the watermark embedding of this step is called relational embedding.

和的离散余弦变换直流系数按照量化步长S分别进行量化，分别得到量化值Q_L，i和Q_R，i， $Q_{L,i}=\mathrm{floor}\left(\frac{{\mathrm{DC}}_{L,i}}{S}\right),$ $Q_{R,i}=\mathrm{floor}\left(\frac{{\mathrm{DC}}_{R,i}}{S}\right),$ 其中，DC_L，i表示

的离散余弦变换直流系数，DC_R，i表示

的离散余弦变换直流系数，S为量化步长，量化步长S的值的选取依据为水印信息在载体图像中的隐藏性和鲁棒性的折中，在此取S＝10，floor(x)为计算得到小于等于x的最大整数函数；定义

的离散余弦变换直流系数的奇偶性为其量化值Q_L，i的奇偶性，定义

的离散余弦变换直流系数的奇偶性为其量化值Q_R，i的奇偶性，判断量化值Q_L，i和量化值Q_R，i是否同为偶数或同为奇数，如果是，则认为

和

的离散余弦变换直流系数的奇偶性一致，否则，认为

和的离散余弦变换直流系数的奇偶性不一致。③-6, right

and The discrete cosine transform DC coefficients are respectively quantized according to the quantization step size S, and the quantized values Q _{L, i} and Q _{R, i} are respectively obtained, $Q_{L,i}=\mathrm{floor}\left(\frac{{\mathrm{DC}}_{L,i}}{S}\right),$ $Q_{R,i}=\mathrm{floor}\left(\frac{{\mathrm{DC}}_{R,i}}{S}\right),$ Among them, DC _{L, i} represent

The discrete cosine transform DC coefficient, DC _{R, i} represents

DC coefficient of discrete cosine transform of , S is the quantization step size, and the selection basis of the value of the quantization step size S is the compromise between the concealment and robustness of the watermark information in the carrier image, here S=10, floor(x ) is calculated to be less than or equal to the largest integer function of x; definition

The parity of the discrete cosine transform DC coefficient is its quantization value Q _{L, the parity of i} , defined

The parity of the discrete cosine transform DC coefficient is the parity of the quantized value Q _{R, i} , and judges whether the quantized value Q _{L, i} and the quantized value Q _{R, i} are both even or odd, and if so, it is considered

and

The parity of DC coefficients of discrete cosine transform is consistent, otherwise, consider

and The parity of the DC coefficients of the discrete cosine transform is inconsistent.

中的第i个二值像素点的像素值w_i是否为“1”，如果是，则修改当前图像块对中的任一个图像块的离散余弦变换直流系数，使修改后的

和

的离散余弦变换直流系数的奇偶性一致，并标记当前图像块对的水印嵌入方式为10，否则，修改当前图像块对

中的任一个图像块的离散余弦变换直流系数，使修改后的

和

的离散余弦变换直流系数的奇偶性不一致，并标记当前图像块对

的水印嵌入方式为10。Judging the binary watermarked image after scrambling

Whether the pixel value w _i of the i-th binary pixel in is "1", if yes, modify the current image block pair The discrete cosine transform DC coefficients of any image block in , so that the modified

and

The parity of the DC coefficients of the discrete cosine transform is consistent, and the current image block pair is marked The watermark embedding mode of is 10, otherwise, modify the current image block pair

The discrete cosine transform DC coefficients of any image block in , so that the modified

and

The parity of the DC coefficients of the discrete cosine transform is inconsistent, and the current image block pair is marked

The watermark embedding method of is 10.

中的任一个图像块的离散余弦变换直流系数时，选择

和

中离散余弦变换直流系数修改幅度小的那个图像块的离散余弦变换直流系数进行修改。In this embodiment, when modifying the current image block pair

When the discrete cosine transform DC coefficient of any image block in , choose

and

The discrete cosine transform direct current coefficient of the image block whose modification range is small among the discrete cosine transform direct current coefficients is modified.

Here, the watermark embedding in this step is called quantized embedding.

③-7. Determine whether i≤m×m is true, if it is true, add 1 to the value of i, use the next image block pair in the original stereoscopic video image S _org as the current image block pair, and repeat step ③-2 Proceed to step ③-7, and end the watermark embedding process until the m×m image block pairs are processed.

④. The size of the binary watermark image, the watermark embedding method and the quantization step S are used as the key, and the watermark embedding end transmits the key to the watermark extraction end.

Figure 3a and Figure 3b show the left-viewpoint image and right-viewpoint image respectively after embedding the binary watermark image shown in Figure 2b. After embedding the watermark, the peak signal-to-noise ratio PSNR of the left view image and the right view image of the "puppy" stereoscopic video image are 49.34dB and 51.34dB respectively, indicating that the stereoscopic video image after embedding the watermark is very similar to the original stereoscopic video image, and the watermark information The concealment is very good.

Embodiment two:

This embodiment proposes a blind watermark extraction method for a stereoscopic video image embedded with a watermark using the blind watermark embedding method given in Embodiment 1, which mainly includes the following steps:

待检测的立体视频图像S_wat的左视点图像L_wat中的第j个图像块

和待检测的立体视频图像S_wat的右视点图像R_wat中的第j个图像块

(即左视点图像L_wat和右视点图像R_wat中同一坐标位置的两个图像块)构成待检测的立体视频图像S_wat中的第j个图像块对，记为

其中，M表示待检测的立体视频图像S_wat的宽，N表示待检测的立体视频图像S_wat的高，1≤j≤((M×N)/(n×n))。① At the watermark extraction end, let S _wat be the stereoscopic video image to be detected _whose size is M×N. The stereoscopic video image S _wat to be detected includes left viewpoint image and right viewpoint image. The left viewpoint image is L _wat , and the right viewpoint image of the stereoscopic video image S _wat to be detected is R _wat , and the left viewpoint image L _wat of the stereoscopic video image S _wat to be detected and the right side of the stereoscopic video image S _wat to be detected are The viewpoint image R _wat is divided into (M×N)/(n×n) non-overlapping image blocks with a size _of n×n, and the j- _th image blocks are denoted as Denote the jth image block in the right viewpoint image R _wat of the stereoscopic video image S _wat to be detected as

The jth image block in the left viewpoint image L _wat of the stereoscopic video image S _wat to be detected

and the jth image block in the right view image R _wat of the stereoscopic video image S _wat to be detected

(that is, two image blocks at the same coordinate position in the left viewpoint image L _wat and the right viewpoint image R _wat ) constitute the jth image block pair in the stereoscopic video image S _wat to be detected, denoted as

Wherein, M represents the width of the stereo video image S _wat to be detected, N represents the height of the stereo video image S _wat to be detected, 1≤j≤((M×N)/(n×n)).

In this embodiment, the size of the stereoscopic video image S _wat to be detected is consistent with the size of the original stereoscopic video image at the watermark embedding end, and the size of the image blocks is also consistent with the size of the image blocks divided by the watermark embedding end.

②The watermark is extracted from each image block pair in the stereoscopic video image S _wat to be detected in turn, and the specific process is as follows:

为当前图像块对。②-1, define the jth image block pair in the stereoscopic video image S _wat to be detected currently being processed

is the current image block pair.

中的

分别进行离散余弦变换和二级离散小波变换，对当前图像块对

中的

分别进行离散余弦变换和二级离散小波变换。②-2. For the current image block pair

middle

Discrete cosine transform and two-level discrete wavelet transform are performed respectively, and the current image block pair

middle

Discrete cosine transform and two-stage discrete wavelet transform are carried out respectively.

的内部关系，判断

的离散余弦变换直流系数是否大于

的二级离散小波变换的设定位置的低频系数，如果是，则置

的内部关系Intra-L′为“1”，否则，置

的内部关系Intra-L′为“0”；令Intra-R′表示的内部关系，判断

的离散余弦变换直流系数是否大于

的二级离散小波变换的设定位置的低频系数，如果是，则置

的内部关系Intra-R′为“1”，否则，置

的内部关系Intra-R′为“0”；令Intra′表示的内部关系，判断

的内部关系Intra-L′和

的内部关系Intra-R′是否均为“1”，如果是，则置

的内部关系Intra′为“1”，否则，置

的内部关系Intra′为“0”；令Inter′表示

的块间关系，判断

的离散余弦变换直流系数是否大于

的离散余弦变换直流系数，如果是，则置

的块间关系Inter′为“1”，否则，置

的块间关系Inter′为“0”。②-3. Let Intra-L′ express

internal relationship, judgment

Is the discrete cosine transform DC coefficient greater than

The low-frequency coefficient of the set position of the second-level discrete wavelet transform, if it is, then set

The internal relationship Intra-L′ of is "1", otherwise, set

The internal relationship Intra-L′ of is “0”; let Intra-R′ represent internal relationship, judgment

Is the discrete cosine transform DC coefficient greater than

The low-frequency coefficient of the set position of the second-level discrete wavelet transform, if it is, then set

The internal relation Intra-R′ of is "1", otherwise, set

The internal relation Intra-R′ of is “0”; let Intra′ represent internal relationship, judgment

The internal relations Intra-L′ and

Whether the internal relations Intra-R′ of are all "1", if yes, then set

The internal relation Intra' of is "1", otherwise, set

The internal relationship Intra' is "0"; let Inter' represent

The inter-block relationship, judging

Is the discrete cosine transform DC coefficient greater than

DC coefficient of discrete cosine transform, if yes, then set

The inter-block relationship Inter' is "1", otherwise, set

The inter-block relation Inter' is "0".

的二级离散小波变换的设定位置的低频系数与

的二级离散小波变换的设定位置的低频系数为各自图像块的二级离散小波变换的相同位置的低频系数。在本实施例中，上述设定位置的低频系数为图像块的二级离散小波变换的任一位置的低频系数，如此处

的二级离散小波变换的低频系数和

的二级离散小波变换的低频系数选取的是各自图像块的低频子带的第0行第1列的低频系数。the above

The low-frequency coefficients of the set position of the second-order discrete wavelet transform with

The low-frequency coefficients at the set position of the second-level discrete wavelet transform are the low-frequency coefficients at the same position of the second-level discrete wavelet transform of the respective image blocks. In this embodiment, the low-frequency coefficient at the above-mentioned setting position is the low-frequency coefficient at any position of the second-level discrete wavelet transform of the image block, as shown here

The low-frequency coefficients of the second-order discrete wavelet transform and

The low-frequency coefficients of the second-level discrete wavelet transform are selected from the low-frequency coefficients in row 0 and column 1 of the low-frequency sub-bands of the respective image blocks.

和

的离散余弦变换直流系数按照量化步长S分别进行量化，分别得到量化值Q′_L，j和Q′_R，j。在此， $Q_{L,j}^{\′}=\mathrm{floor}\left(\frac{{{\mathrm{DC}}^{\′}}_{L,j}}{S}\right),$ $Q_{R,j}^{\′}=\mathrm{floor}\left(\frac{{{\mathrm{DC}}^{\′}}_{R,j}}{S}\right),$ 其中，DC′_L，j表示

的离散余弦变换直流系数，DC′_R，j表示的离散余弦变换直流系数，S为量化步长，floor(x)为计算得到小于等于x的最大整数函数。②-4, right

and

The discrete cosine transform DC coefficients are quantized according to the quantization step size S, and the quantized values Q′ _{L, j} and Q′ _{R, j} are respectively obtained. here, $Q_{L,j}^{\′}=\mathrm{floor}\left(\frac{{{\mathrm{DC}}^{\′}}_{L,j}}{S}\right),$ $Q_{R,j}^{\′}=\mathrm{floor}\left(\frac{{{\mathrm{DC}}^{\′}}_{R,j}}{S}\right),$ Among them, DC′ _{L, j} represents

The DC coefficient of discrete cosine transform, DC′ _{R, j} represents The discrete cosine transform DC coefficient of , S is the quantization step size, and floor(x) is the largest integer function less than or equal to x calculated.

的水印嵌入方式是否为00，如果不是，则执行步骤②-6，否则，再判断

的内部关系Intra-L′和

的内部关系Intra-R′是否均为“1”，如果均为“1”，则从当前图像块对

中提取出水印“1”，如果均为“0”或其中一个为“0”，则从当前图像块对

中提取出水印“0”，然后执行步骤②-8。②-5. Judging the current image block pair

Whether the embedding method of the watermark is 00, if not, go to step ②-6, otherwise, judge again

The internal relations Intra-L′ and

Whether the internal relationship Intra-R′ of is all "1", if they are all "1", then from the current image block to

Extract the watermark "1" from the current image block, if they are all "0" or one of them is "0", then the

Extract the watermark "0" from the data, and then perform steps ②-8.

中提取出水印“1”，如果为“0”，则从当前图像块对

中提取出水印“0”，然后执行步骤②-8。②-6. Judging the current image block pair Whether the embedding method of the watermark is 01, if not, execute steps ②-7, otherwise, judge the current image block pair Is the inter-block relationship Inter' of "1", if it is "1", then from the current image block to

Extract the watermark "1" from the watermark, if it is "0", then from the current image block pair

Extract the watermark "0" from the data, and then perform steps ②-8.

的水印嵌入方式为10时，判断Q′_L，j和Q′_R，j是否均为奇数或者均为偶数，如果是，则从当前图像块对中提取出水印“1”，否则，从当前图像块对

中提取出水印“0”，然后执行步骤②-8。②-7. When the current image block is

When the watermark embedding method is 10, it is judged whether Q′ _{L, j} and Q′ _{R, j} are all odd numbers or even numbers, and if so, from the current image block to Extract the watermark "1" from the current image block, otherwise, from the current image block pair

Extract the watermark "0" from the data, and then perform steps ②-8.

②-8. Determine whether j≤m×m is true. If it is true, add 1 to the j value, use the next image block pair in the stereoscopic video image S _wat to be detected as the current image block pair, and repeat step ② -2 to step ②-8, until the m×m image block pairs are processed and the watermark extraction process ends.

③ Perform inverse scrambling transformation on the extracted watermark image to obtain the final extracted binary watermark image. In this embodiment, since L times of Arnold scrambling transformations are used, the period Z of the Arnold scrambling transformation can be determined according to the size of the binary watermark image transmitted from the watermark embedding end to the watermark extraction end, and then all watermarks obtained by extraction are ( Z-L) Arnold scrambling transformation to obtain the final extracted binary watermark image.

来计算提取得到的二值水印图像和原始二值水印图像的相似度，0≤HC≤1，其中w(s，t)和

分别表示原始二值水印图像和提取出的二值水印图像坐标位置为(s，t)处的像素值，

表示异或运算符。HC越大，表示提取得到的二值水印图像越接近原始二值水印图像，HC越小，表示提取得到的二值水印图像与原始二值水印图像相差比较大。通过HC值的大小可以评价水印嵌入算法的鲁棒性。在没有攻击的情况下，本实施例从图3a和图3b构成的立体视频图像中提取出的二值水印图像的归一化相似值HC为1，即和原始二值水印图像完全相同。Figure 4a and Figure 4b respectively show all the watermarks extracted from the stereoscopic video images formed in Figure 3a and Figure 3b when there is no attack and the final extracted binary watermark image. In this embodiment, using the normalized similarity value HC,

To calculate the similarity between the extracted binary watermark image and the original binary watermark image, 0≤HC≤1, where w(s, t) and

Represent the original binary watermark image and the pixel value at the coordinate position (s, t) of the extracted binary watermark image,

Represents the XOR operator. The larger the HC, the closer the extracted binary watermark image is to the original binary watermark image, and the smaller the HC, the larger the difference between the extracted binary watermark image and the original binary watermark image. The robustness of the watermark embedding algorithm can be evaluated by the value of HC. In the case of no attack, the normalized similarity value HC of the binary watermark image extracted from the stereo video image composed of Fig. 3a and Fig. 3b in this embodiment is 1, which is exactly the same as the original binary watermark image.

Further, the obtained stereoscopic video images embedded with watermarks are subjected to distortion processing, including JPEG compression under different quality factors, Gaussian filtering with different window sizes, median filtering, mean filtering, and cutting of different degrees of positions. Fig. 5a, Fig. 5b, Fig. 5c, Fig. 5d and Fig. 5e respectively provide the binary watermark images extracted from the JPEG-compressed stereoscopic video images whose quality factors Q are 90, 70, 50, 30, 20 respectively; Fig. 6a, Fig. 6b, Fig. 6c, Fig. 6d and Fig. 6e respectively show Gaussian filtering (window 3×3), median filtering (window 3×3), median filtering (window 5×5), mean filtering (window 3×3) and mean filtering (window 5×5) of the binary watermark image extracted from the stereoscopic video image; Fig. 7a, Fig. 7b, Fig. 7c, Fig. 7d and Fig. 7e respectively show the The binary watermark image extracted from the stereoscopic video image after cutting 1/4, cutting the top left end 1/8, cutting the middle 1/4, cutting the middle 1/8, and cutting 1/2 the top left end. Table 1 shows the normalized similarity value HC of the binary watermark image extracted after the above-mentioned various distortion treatments. From Figure 5a to Figure 7e and Table 1, it can be seen that under different types and degrees of distortion attacks, the watermark embedded by the method of the present invention has better identification, which shows that the embedding and extraction method of the present invention has better robustness .

Table 1 The normalized similarity value HC of watermark detection under different types and degrees of distortion attacks

JPEG compression Q=90 JPEG compression Q=70 JPEG compression Q=50 JPEG compression Q=30 JPEG compression Q=20 HC＝0.9258 HC＝0.9063 HC＝0.8601 HC＝0.7324 HC＝0.7048 Gaussian filter odd window Median filtering 3×3 Median filter 5×5 Mean filtering 3×3 Mean filtering 5×5

HC＝0.9932 HC＝0.8965 HC＝0.8091 HC＝0.9216 HC＝0.7986 Cut 1/4 from the top left end Cut 1/8 from the top left end Middle cut 1/4 Middle cut 1/8 Cut 1/2 from the top left end HC＝0.9198 HC＝0.9819 HC＝0.8369 HC＝0.9216 HC＝0.8438

Claims (5)

1. the blind watermark embedding method of a stereoscopic video images is characterized in that may further comprise the steps:

1. at the watermark built-in end, make S _OrgFor being of a size of the original stereoscopic video images of M * N, original stereoscopic video images S _OrgComprise left visual point image and right visual point image, remember original stereoscopic video images S _OrgLeft visual point image be L _Org, remember original stereoscopic video images S _OrgRight visual point image be R _Org, with original stereoscopic video images S _OrgLeft visual point image L _OrgWith original stereoscopic video images S _OrgRight visual point image R _OrgBe divided into respectively (M * N)/(n * n) individual non-overlapping copies is of a size of the image block of n * n, with original stereoscopic video images S _OrgLeft visual point image L _OrgIn i image block be designated as

With original stereoscopic video images S _OrgRight visual point image R _OrgIn i image block be designated as

Original stereoscopic video images S _OrgLeft visual point image L _OrgIn i image block With original stereoscopic video images S _OrgRight visual point image R _OrgIn i image block Constitute original stereoscopic video images S _OrgIn i image block right, be designated as

Wherein, M represents original stereoscopic video images S _OrgWide, N represents original stereoscopic video images S _OrgHeight, 1≤i≤((M * N)/(n * n));

2. make that W is the binary bitmap that is of a size of m * m, binary bitmap W is carried out the scramble conversion, obtain the binary bitmap behind the scramble, be designated as

Wherein, m * m≤((M * N)/(n * n));

3. successively to original stereoscopic video images S _OrgIn each image block to adding watermark, its detailed process is as follows:

3.-1, the current original stereoscopic video images S that is handling of definition _OrgIn i image block right

For current image block right;

3.-2, right to current image block

In

Carry out discrete cosine transform and secondary wavelet transform respectively, right to current image block

In

Carry out discrete cosine transform and secondary wavelet transform respectively;

3.-3, make Intra-L represent

Internal relations, judge

The discrete cosine transform DC coefficient whether greater than

The low frequency coefficient of desired location of secondary wavelet transform, if then put

Internal relations Intra-L be " 1 ", otherwise, put

Internal relations Intra-L be " 0 "; Make Intra-R represent

Internal relations, judge The discrete cosine transform DC coefficient whether greater than The low frequency coefficient of desired location of secondary wavelet transform, if then put

Internal relations Intra-R be " 1 ", otherwise, put

Internal relations Intra-R be " 0 "; Make Intra represent

Internal relations, judge

Internal relations Intra-L and

Internal relations Intra-R whether be " 1 ", if then put Internal relations Intra be " 1 ", otherwise, put

Internal relations Intra be " 0 "; Make Inter represent

Interblock relation, judge

The discrete cosine transform DC coefficient whether greater than The discrete cosine transform DC coefficient, if then put

Interblock relations I nter be " 1 ", otherwise, put

Interblock relations I nter be " 0 ";

Above-mentioned

The secondary wavelet transform desired location low frequency coefficient with

The low frequency coefficient of desired location of secondary wavelet transform be the low frequency coefficient of same position of the secondary wavelet transform of image block separately;

3.-4, judge that current image block is right

The value of internal relations Intra and the binary bitmap behind the scramble

In the pixel value w of i binarized pixel point _iWhether identical, if identical, then do not revise

With The discrete cosine transform DC coefficient, and the mark current image block is right

The watermark embedded mode be 00, execution in step is 3.-7 then; Otherwise execution in step 3.-5;

3.-5, judge that current image block is right

The value of interblock relations I nter and the binary bitmap behind the scramble

In the pixel value w of i binarized pixel point _iWhether identical, if identical, then do not revise

With The discrete cosine transform DC coefficient, and the mark current image block is right

The watermark embedded mode be 01, execution in step is 3.-7 then; Otherwise execution in step 3.-6;

3.-6, right

With

The discrete cosine transform DC coefficient quantize respectively according to quantization step S, obtain quantized value Q respectively _{L, i}And Q _{R, i}, $Q_{L,i}=\mathrm{floor}\left(\frac{{\mathrm{DC}}_{L,i}}{S}\right),$ $Q_{R,i}=\mathrm{floor}\left(\frac{{\mathrm{DC}}_{R,i}}{S}\right),$ Wherein, DC _{L, i}Expression

The discrete cosine transform DC coefficient, DC _{R, i}Expression

The discrete cosine transform DC coefficient, floor (x) is for calculating greatest integer function smaller or equal to x; Definition

The parity of discrete cosine transform DC coefficient be its quantized value Q _{L, i}Parity, the definition

The parity of discrete cosine transform DC coefficient be its quantized value Q _{R, i}Parity, judge quantized value Q _{L, i}With quantized value Q _{R, i}Whether be all even number or be all odd number, if then think

With

The parity unanimity of discrete cosine transform DC coefficient, otherwise, think With

The parity of discrete cosine transform DC coefficient inconsistent;

Binary bitmap behind the judgement scramble

In the pixel value w of i binarized pixel point _iWhether be " 1 ", if it is right then to revise current image block

In the discrete cosine transform DC coefficient of any image block, make amended

With

The parity unanimity of discrete cosine transform DC coefficient, and the mark current image block is right

The watermark embedded mode be 10, otherwise it is right to revise current image block

In the discrete cosine transform DC coefficient of any image block, make amended With

The parity of discrete cosine transform DC coefficient inconsistent, and the mark current image block is right

The watermark embedded mode be 10;

3.-7, judge whether i≤m * m sets up, if establishment then adds 1 with the i value, with original stereoscopic video images S _OrgIn next image block to right as current image block, and repeated execution of steps 3.-2 to step 3.-7, until handling m * m image block to finishing watermark embed process;

4., the size with binary bitmap, watermark embedded mode and quantization step S be as key, the watermark built-in end extracts end with cipher key delivery to watermark.

2. the blind watermark embedding method of a kind of stereoscopic video images according to claim 1 is characterized in that the low frequency coefficient of desired location during described step 3.-3 is the low frequency coefficient of arbitrary position of the secondary wavelet transform of image block.

3. the blind watermark embedding method of a kind of stereoscopic video images according to claim 1 and 2 is characterized in that during described step 3.-6 in that to revise current image block right

In the discrete cosine transform DC coefficient of any image block the time, select

With

Middle discrete cosine transform DC coefficient is revised the discrete cosine transform DC coefficient of that little image block of amplitude and is made amendment.

4. blind watermark extracting method that uses the watermarked stereoscopic video images of the described blind watermark embedding method of claim 1 is characterized in that may further comprise the steps:

1. at the watermark extracting end, make S _WatFor being of a size of the stereoscopic video images to be detected of M * N, stereoscopic video images S to be detected _WatComprise left visual point image and right visual point image, remember stereoscopic video images S to be detected _WatLeft visual point image be L _Wat, remember stereoscopic video images S to be detected _WatRight visual point image be R _Wat, with stereoscopic video images S to be detected _WatLeft visual point image L _WatWith stereoscopic video images S to be detected _WatRight visual point image R _WatBe divided into respectively (M * N)/(n * n) individual non-overlapping copies is of a size of the image block of n * n, with stereoscopic video images S to be detected _WatLeft visual point image L _WatIn j image block be designated as

With stereoscopic video images S to be detected _WatRight visual point image R _WatIn j image block be designated as

Stereoscopic video images S to be detected _WatLeft visual point image L _WatIn j image block

With stereoscopic video images S to be detected _WatRight visual point image R _WatIn j image block Constitute stereoscopic video images S to be detected _WatIn j image block right, be designated as

Wherein, M represents stereoscopic video images S to be detected _WatWide, N represents stereoscopic video images S to be detected _WatHeight, 1≤j≤((M * N)/(n * n));

2. successively to stereoscopic video images S to be detected _WatIn each image block to extracting watermark, its detailed process is as follows:

2.-1, the current stereoscopic video images S to be detected that is handling of definition _WatIn j image block right

For current image block right;

2.-2, right to current image block

In Carry out discrete cosine transform and secondary wavelet transform respectively, right to current image block

In

Carry out discrete cosine transform and secondary wavelet transform respectively;

2.-3, make Intra-L ' expression

Internal relations, judge

The discrete cosine transform DC coefficient whether greater than

The low frequency coefficient of desired location of secondary wavelet transform, if then put

Internal relations Intra-L ' be " 1 ", otherwise, put Internal relations Intra-L ' be " 0 "; Make Intra-R ' expression

Internal relations, judge

The discrete cosine transform DC coefficient whether greater than

The low frequency coefficient of desired location of secondary wavelet transform, if then put Internal relations Intra-R ' be " 1 ", otherwise, put

Internal relations Intra-R ' be " 0 "; Make Intra ' expression

Internal relations, judge

Internal relations Intra-L ' and Internal relations Intra-R ' whether be " 1 ", if then put

Internal relations Intra ' be " 1 ", otherwise, put Internal relations Intra ' be " 0 "; Make Inter ' expression

Interblock relation, judge

The discrete cosine transform DC coefficient whether greater than The discrete cosine transform DC coefficient, if then put

Interblock relations I nter ' be " 1 ", otherwise, put

Interblock relations I nter ' be " 0 ";

Above-mentioned

The secondary wavelet transform desired location low frequency coefficient with

The low frequency coefficient of desired location of secondary wavelet transform be the low frequency coefficient of same position of the secondary wavelet transform of image block separately;

2.-4, right

With

The discrete cosine transform DC coefficient quantize respectively according to quantization step S, obtain quantized value Q ' respectively _{L, j}And Q ' _{R, j}, $Q_{L,j}^{\′}=\mathrm{floor}\left(\frac{{{\mathrm{DC}}^{\′}}_{L,j}}{S}\right),$ $Q_{R,j}^{\′}=\mathrm{floor}\left(\frac{{{\mathrm{DC}}^{\′}}_{R,j}}{S}\right),$ Wherein, DC ' _{L, j}Expression

The discrete cosine transform DC coefficient, DC ' _{R, j}Expression

The discrete cosine transform DC coefficient, S is a quantization step, floor (x) is for calculating greatest integer function smaller or equal to x;

2.-5, judge that current image block is right

The watermark embedded mode whether be 00, if not, then execution in step 2.-6, otherwise, judge again

Internal relations Intra-L ' and

Internal relations Intra-R ' whether be " 1 ", if be " 1 ", then right from current image block

In extract watermark " 1 ", be " 0 " if be " 0 " or one of them, then right from current image block

In extract watermark " 0 ", execution in step is 2.-8 then;

2.-6, judge that current image block is right

The watermark embedded mode whether be 01, if not, then execution in step 2.-7, otherwise, judge that again current image block is right

Interblock relations I nter ' whether be " 1 ", if be " 1 ", then right from current image block

In extract watermark " 1 ", if be " 0 ", then right from current image block

In extract watermark " 0 ", execution in step is 2.-8 then;

2.-7, judge Q ' _{L, j}And Q ' _{R, j}Whether be odd number or be even number, if, then right from current image block In extract watermark " 1 ", otherwise, right from current image block

In extract watermark " 0 ", execution in step is 2.-8 then;

2.-8, judge whether j≤m * m sets up, if establishment then adds 1 with the j value, with stereoscopic video images S to be detected _WatIn next image block to right as current image block, and repeated execution of steps 2.-2 to step 2.-8, until handling m * m image block to finishing watermark extraction process;

3. the watermarking images that extraction the is obtained random conversion that is inverted, the binary bitmap of finally being extracted.

5. the blind watermark extracting method of a kind of stereoscopic video images according to claim 4 is characterized in that the low frequency coefficient of desired location during described step 2.-3 is the low frequency coefficient of arbitrary position of the secondary wavelet transform of image block.

Images