CN1207892C - Method and device for inserting digital image into electronic water-print and detecting method - Google Patents

Abstract

An image is divided into a plurality of blocks, with DCT converter (103), electronic watermark data inserter (104), and IDCT converter (107) different electronic watermark data are embedded in each block, at this moment refer to for An electronic watermark data insertion table (105) and an electronic watermark data table (106) for inserting different electronic watermarks into different image blocks are determined.

Description

Method and device for inserting electronic watermark into digital image and detecting it

technical field

The present invention relates to the field of digital processing, and more specifically relates to the technology of embedding or inserting identification data (electronic watermark data) with specific information into digital images.

Background technique

In recent years, illegal duplication of digital images has become a problem. In order to prevent such illegal duplication, it has been considered to encode digital image data, and only a reproduction system having a legitimate key can reproduce the decoded digital data. However, once the password has been decrypted, the system cannot prevent subsequent copying.

Therefore, in order to prevent illegal use and copying of digital images, a method has been considered in which specific information (refer to electronic watermark data described below) is embedded in a data image. For such watermark data for digital images, two types of electronic watermark data such as visible and invisible can be used.

Visible electronic watermark data becomes intuitive and visible by synthesizing specific words or codes to images. Although the insertion of such visible electronic watermark data will cause image quality degradation, it can intuitively prevent illegal conversion of the user's digital image.

An example of visible electronic watermark data is disclosed in Japanese Patent Laid-Open No. Hei-8-241403. In this example, when the visible electronic watermark data is synthesized into the original image, without changing the color components, only the luminance components of the pixels corresponding to the dark parts of the electronic watermark data can be changed by a conversion value. Electronic watermark data is synthesized into the original image. The conversion value is determined by the color components, random numbers and pixel values of the electronic watermark data.

On the contrary, in the case of invisible electronic watermark data, when the electronic watermark data is embedded in the image and prevents the image quality from degrading, there is almost no damage to the image quality derived from the inserted electronic watermark data, thus embedding The electronic watermark data of the digital watermark is not intuitive and visual. If information identifying the author as electronic watermark data is embedded in an image, the author can be identified by detecting this electronic watermark data even after illegal copying has been performed. In addition, if information prohibiting copying is embedded as electronic watermark data, when the copying device detects the copying prohibition information, for example, the copying device can notify the user that the image is copy-prohibited data, or activate the copying prohibition mechanism in the copying device to limit the video tape recorder or a copy of a similar device.

As a method of embedding invisible electronic watermark data into digital images, in which specific information as electronic watermark data is embedded in a certain part, while image quality such as the least significant bit (LSB) of a pixel or the like amount) will not have much impact. However, in this method, electronic watermark data can be easily deleted from the image. For example, if low-pass filtering is performed on an image embedded with digitally watermarked data, the information of the LSB of the pixel is lost, so that the digitally watermarked data is deleted without any perceptible corruption. In addition, as an image compression technique that reduces the amount of data by reducing the amount of information in parts that have little influence on image quality, by performing a compression and decompression process on image data with embedded electronic watermark data, it can be used in the The electronic watermark data is deleted when the image quality is less affected.

In addition, another example is disclosed in Japanese Patent Application Laid-Open No. Hei 6-315131. It describes that, when an image is reproduced by utilizing the similarity of images in consecutive frames, an area whose image quality is not damaged even when the area is replaced with an outer area is detected, and by changing Transforms a region's data to embed a particular region within that region. This method reproduces an image by reconstructing the image by first identifying an area in which identification data has been partially embedded by using a loss signal and converting the information, and then correcting the identification area with peripheral data in the identification area.

Also disclosed in Japanese Patent Laid-Open No. Hei 5-30466 is a method in which the frequency of an image signal is converted, and information of a frequency lower than the frequency range of the converted image signal is embedded. This method extracts the raw image signal with a high-pass filter and extracts the embedded identification data with a low-pass filter.

As another example of frequency conversion, in this method, an image is frequency-converted, and after the frequency conversion has been completed, regions with strong frequency components are embedded in the image signal (April 22, 1996, "Nikkei Electronics" No. 660, page 13). This method is to embed the electronic watermark data in an area with strong frequency components, so that the electronic watermark data will not be lost even if processes such as compression and decompression or filtering are performed on the image. As long as the original data is damaged, the electronic watermark data will not be deleted. Furthermore, by employing random numbers according to a normal distribution, interference between digital watermark data can be prevented, and further, by embedding data, degradation of image quality in which digital watermark data has been embedded can be prevented.

The method of embedding electronic watermark data is as follows, the original image is converted into frequency components by DCT (discrete cosine transform) or similar methods, and n data representing high values f(1), f(2), ... are selected in the frequency domain .f(n), select digital watermark data w(1), w(2),...w(n) from a normal distribution with a mean of 0 and a variance of 1, and use the following formula to calculate each i.

F(i)=f(i)+α×|f(i)|×w(i)

Where α is the conversion factor.

By performing inverse DCT transformation on F(i), the image embedded with electronic watermark data is obtained.

The detection of electronic watermark data is performed in the following method. In this detection method, the original image f(i) and the candidate digital watermark data w(i) (where i is 1, 2, . . . n) must be known.

First, the image with electronic watermark data is converted into frequency components by DCT, and the element values of f(1), f(2)...f(n) corresponding to the embedded electronic watermark data in the frequency domain are converted into F(1), F(2), . . . F(n). Use the following formula to calculate and obtain the statistical similarity between w(i) and W(i) through f(i) and F(i).

w(i)=(F(i)-f(i))/f(i)

Next, calculate the statistical similarity of w(i) and W(i) using the vector inner product and the following formula.

C＝W·w/(WD/wD)

Wherein: W=(W(1), W(2), . . . , W(n));

w = (w(1), w(2), ..., w(n));

WD = absolute value of vector W

wD = the absolute value of the vector w; and

· is the inner product of vectors.

When the statistical similarity C is greater than some specific value, it can be predicted that the electronic watermark data may be embedded.

If electronic watermark data is embedded in an image in this way, it is effective when the author of the original image detects a digital image suspected of being an illegal copy. This method requires the original image so that the electronic watermark data can be detected when the author detects the image data suspected of being an illegal copy, however, since the reproduction device does not have the original image, it is not possible in each terminal of the reproduction device Electronic watermark data detected.

Therefore, this method is improved and used for terminal processing, especially for MPEG systems. In this method, an original image is divided into blocks of 8 pixels by 8 pixels, and digital watermark data is embedded in and extracted from the blocks as processing units.

The embedding process of electronic watermark data is performed by the following process. First, after the discrete cosine transform has been performed in the MPEG decoding process in the frequency domain, f(1), f(2),...f(n) are defined sequentially from the lower frequency components in the AC component, and from the mean Choose w(1), w(2), ...w(n) for a normal distribution with 0 and variance 1, and calculate each i with the following formula:

F(i)=f(i)+α×avg(f(i))×w(i)

Where α is the conversion factor, and avg(f(i)) is the partial average of the average of the absolute values of the three adjacent points of f(i).

Then, replace f(i) with F(i) for subsequent MPEG decoding process.

The electronic watermark data is detected by the following procedure. In this detection method, the original image is not needed, but only the candidate value w(i) of the electronic watermark data (where i is 1, 2, 3, ... n) is needed.

In the decompression or decoding process of MPEG, in the frequency domain of the dequantized block, F(1), F(2), . . . F(n) are sequentially determined from the lower frequency components. The average of the absolute values of the three adjacent points F(i-1), F(i), and F(i+1) of F(i) is defined as the partial average value avg(F(i)), and Electronic watermark data W(i) is calculated with the following formula:

W(i)=F(i)/avg(F(i))

And compute the "sum" WF(i) of the image for each i.

Next, use the inner product of the vectors and the following formula to calculate the statistical similarity between w(i) and WF(i).

C＝WF·w/(WFD×wD)

When the statistical similarity is greater than some specific value, it can be considered that the electronic watermark data has been embedded.

In the example disclosed in Japanese Patent Application Laid-Open No. 6-315131, digital watermark information is not embedded in all frames, so that illegal copying cannot be prevented for frames in which the digital watermark is not embedded. In addition, this example assumes that consecutive frames are still images and there is no change in consecutive frames, so that in a movie image including fast moving images, an area in which digital watermark data is embedded cannot be identified, making it impossible to embed digital watermark data.

In addition, in an example disclosed in Japanese Patent Application Laid-Open No. 5-30466, electronic watermark data is embedded in a frequency domain lower than the frequency domain in which the image has been frequency-converted, so that the electronic watermark can be easily converted to the frequency domain using a high-pass filter. The data is deleted.

Furthermore, in an instance where electronic watermark data is embedded in a strong part of the frequency components after the image has been frequency-converted, the electronic watermark data cannot be deleted by a filter or the like. However, when a plurality of electronic watermark data is embedded in one image, there is a problem that if a plurality of electronic watermark data are embedded in the same frequency, the electronic watermark data interfere with each other, thereby reducing its detectability when detected .

Contents of the invention

An object of the present invention is an electronic watermark data inserter and detector in which even when a plurality of electronic watermark data are inserted into an image, they do not interfere with each other and upon detection, the electronic watermark data can be accurately detected.

In the method of inserting identification data (electronic watermark data) according to the present invention, the image of each block of j×k pixels (where j and k are natural numbers) is subjected to frequency conversion, and when the electronic watermark data is inserted into the converted frequency In the composition, after changing the type of electronic watermark data to be inserted, insert electronic watermark data into each block, so that the electronic watermark data is inserted without interfering with each other, and each block is detected . More specifically, an electronic watermark data insertion location table is also provided, which indicates which type of electronic watermark data is inserted into which block of the image, and an applicable electronic watermark data is inserted into the area shown in the insertion area table . When detecting electronic watermark data, an electronic watermark data extraction table similar to that at the time of insertion is used for detecting electronic watermark data.

In the present invention, when a plurality of electronic watermark data is inserted into an image, different electronic watermark data is inserted for each block, thereby inserting the electronic watermark data without mutual interference and damage, also even when embedding a plurality of electronic watermark data When watermarking data, digital watermark data can also be detected correctly.

Description of drawings

Fig. 1 is the block diagram of electronic watermark data insertion device according to the present invention;

Fig. 2 is the block diagram of electronic watermark data detecting device according to the present invention;

FIG. 3 is a schematic diagram of the insertion position of each type of digital watermark data in an image.

Detailed ways

Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

Fig. 1 is a block diagram of an electronic watermark data inserting apparatus when n digital watermark data are inserted into image data.

Referring to FIG. 1, an embodiment of an interpolation device according to the present invention includes a DCT transformer 103 that outputs position information of a block of j×k pixels extracted from an image and performs discrete cosine transform (DCT) on this information, and performs an orthogonal Output data after conversion; an electronic watermark data insertion position table 105, which records the block positions of each type of n types of electronic watermark data to be inserted in the image; an electronic watermark data table 106, which stores n types of electronic watermark data; An electronic watermark data inserter 104, which extracts applicable electronic watermark data from the electronic watermark data table 106 according to the electronic watermark data insertion position table 105 and DCT converter 103, and inserts the electronic watermark data into the data after DCT conversion; And an IDCT converter 107 that performs an inverse discrete cosine transform (IDCT) on the data within the block after DCT conversion, wherein the data is interpolated within the block.

Fig. 2 is a block diagram of a detection device for electronic watermark data when n types of electronic watermark data are inserted into image data.

Referring to FIG. 2, an example of a detection device according to the present invention includes a DCT transformer 203 that outputs position information of a block of j×k pixels taken out from an image and performs discrete cosine transform (DCT) on the information, and after DCT transform Output data; an electronic watermark data extraction position table 205, which stores information representing the position where each of the n types of electronic watermark data is taken out from the image block; and an extracted data table 206, which stores the information of the entire screen The extracted data of n types; an electronic watermark data table 208, which stores n types of electronic watermark data; Extract the electronic watermark data from the frequency data of ×k pixels, and store the extracted electronic watermark data into a predetermined position in the extracted data table 206; The table 206 extracts the mth data (m=1, 2, 3, . . . n), and calculates the statistical similarity between the extracted data and the electronic watermark data.

In the present invention, the contents of the electronic watermark data table and the corresponding number of electronic watermark data location tables must agree with each other on the inserting side and the detecting side. That is, the corresponding contents of the m-th (m=1, 2, 3, . . . n) digital watermark data tables on the insertion and detection sides must correspond to each other. In addition, the content of the digital watermark data insertion position table 105 on the insertion device side must be consistent with the content of the digital watermark data extraction position table 205 on the detection device side.

Referring to Fig. 1, the operation of the electronic watermark data inserting apparatus will be described.

A DCT transformer 103 takes out block data 102 with a size of 8×8 pixels from the original image to perform DCT transformation. The electronic watermark data inserter 104 takes out a plurality of electronic watermark data to be inserted from the electronic watermark data insertion position table 105 on the basis of the block position data output by the DCT transformer 103 . Then, the digital watermark data inserter 104 takes out the digital watermark data whose number is equal to the number taken out by the digital watermark data insertion position table 105 from the digital watermark data table 106, so as to insert the digital watermark data into the DCT-transformed data output by the DCT converter 103 in the data.

The IDCT transformer 107 performs IDCT transformation on the data output by the electronic watermark data inserter 104, and stores the data in the same location 109 as that where the DCT transformer 103 has taken out the block data in the image storage area 108 where the data is to be inserted. same location.

The above-described operations are performed on all the blocks specified by the electronic watermark data insertion position table 105 . In addition, the original image data 101 in the block not specified by the electronic watermark data insertion position table is directly copied into the memory which temporarily stores the image data 108 through a processing path not shown in FIG. 1 .

Next, referring to FIG. 2, the operation of the electronic watermark data detection apparatus will be described.

The DCT transformer 203 takes out block data 202 with a size of 8×8 pixels from the image 201 to perform DCT transformation. The electronic watermark data extractor 204 takes out a plurality of electronic watermark data to be extracted from the electronic watermark data extraction position table 205 on the basis of the 8×8 block position data output by the DCT converter 203 . Then, the digital watermark data extractor 204 extracts the digital watermark extraction data from the DCT-transformed frequency data output by the DCT converter 203, and stores the data in the data corresponding to the digital watermark data extracted from the extracted data storage area 206. The location of the electronic watermark data obtained in Table 205.

The above operation is performed on all the blocks specified by the digital watermark data extraction position table 205 .

After one screen of extracted data is stored in the extracted data storage area 206, the electronic watermark data detector 207 takes out the corresponding extracted data and electronic watermark data from the extracted data storage area and the electronic watermark data table 208, and calculates the statistical similarity to output the result.

When there are n types of digital watermark data, the digital watermark data detector 207 repeats the above operations n times, and calculates the statistical similarity between all digital watermark data and the extracted data to output the result 209 .

Fig. 3 is a schematic diagram of the electronic watermark data insertion position table, wherein the insertion position table indicates the position where each electronic watermark data is inserted and the value of the detection position table when four types of electronic watermark data are inserted. The image data of the entire screen is divided into blocks having a size of j*k shown in FIG. 3, and stored is an ordinal number representing electronic watermark data to be inserted for each block.

The example in Figure 3 points out that when the upper left of the image is set as the original value, the first digital watermark data is inserted in the block whose column number is 1 and row number is 1, and the first digital watermark data is inserted in a block displaced by one block in the horizontal direction. 2. Electronic watermark data.

Referring to FIG. 1 to FIG. 3, the operation of the electronic watermark data inserting apparatus will be described more specifically.

The DCT converter 103 takes out block data 102 of size j×k from the position where the row number and column number of the original image 101 are both 1, performs DCT transformation, and outputs data representing row and column numbers and frequency data after DCT transformation . The digital watermark data inserter 104 fetches the digital watermark data numbered 1 to be inserted from the digital watermark data insertion position table 105 on the basis of the position data of row number 1 and column number 1 of the block output by the DCT converter 103 .

Then, the electronic watermark data inserter 104 takes out the first electronic watermark data corresponding to the electronic watermark data labeled 1 from the electronic watermark data table 106, and inserts the first electronic watermark data into the DCT outputted by the DCT converter 103. In the transformed frequency data. IDCT converter 107 performs IDCT conversion on the data output by electronic watermark data inserter 104, and stores the data in the position where the row number and column number are 1, which is the same as the position where the DCT converter has extracted data from the area 108 Likewise, an image into which digital watermark data has been inserted is stored in the area 108 .

Then, similar operations are performed on the block data whose row number and column number are both 1. The above-described operation is performed for all the blocks specified by the electronic watermark data insertion position table 105 (in the example of FIG. 3, for all data in the image data).

Next, with reference to FIG. 2 and FIG. 3 , the operation of the digital watermark data detection device will be described in more detail.

The DCT converter 203 extracts block data of pixels with a size of 8×8 from the position where the row number and column number of the image 201 are both 1 and performs DCT transformation, and outputs row and column number data and frequency data after DCT transformation. The digital watermark data extractor 204 obtains the digital watermark data to be extracted with the label 1 from the digital watermark data extraction table 205 on the basis of the position data of the block whose row number and column number are both 1 output by the DCT converter 203 .

Then, the digital watermark data extractor 204 extracts the digital watermark data from the DCT-transformed frequency data outputted by the DCT converter 203, and stores it in the digital watermark data corresponding to the number 1 in the extracted data storage area 206. an extraction area.

Next, similar processing is performed on the block data whose row number and column number are both 1. The above-mentioned operation is performed on all block data specified by the digital watermark data extraction position table (in the example of FIG. 3 , all blocks in the image data).

After storing the extracted data of the whole screen into the extracted data storage area, the electronic watermark data detector 207 takes out the first extracted data from the extracted data storage area 206 and the first electronic watermark data from the electronic watermark data table 208, and calculates the statistical similarity degree to output the result.

Next, the electronic watermark data detector 207 retrieves the second extracted data from the extracted data storage area 206 and the second electronic watermark data from the electronic watermark data table 208, and calculates the statistical similarity to output the result.

In a similar manner, similar processing is also performed on the third extracted data and the third electronic watermark data and the fourth extracted data and the fourth electronic watermark data.

As described above, the present invention divides an image into block units of j×k pixels and changes the electronic watermark data to be inserted for each block, thereby preventing multiple The problem of mutual interference of electronic watermark data.

For those skilled in the art, various changes and modifications to the present invention are within the scope defined by the appended claims.

Claims (2)

1. one kind is inserted device in the digital picture with electronic watermark data, it is characterized in that comprising:

A quadrature transformer, it takes out the image block that is of a size of j * k from digital picture, and wherein j and k are natural number, and described image block is carried out orthogonal transform, and the positional information of the image block of output taking-up and the data after the orthogonal transform;

An electronic watermark data extracts the position table, the positional information that each class of the electronic watermark data of its storage representation n class is taken out from image block, and wherein n is a positive integer;

An extracted data table, the image block data that comprises n electron-like watermark data of the whole screen of its storage;

One electronic watermark data table, its storage n electron-like watermark data;

One electronic watermark data withdrawal device, it extracts the position according to the positional information by described quadrature converter output from described electronic watermark data and show the positional information that each class of the electronic watermark data of extraction expression n class is taken out from image block, extract the frequency data that comprise electronic watermark data from frequency data by the piece of described quadrature transformer output, and storing frequency data in the described extracted data table precalculated position, described extracted data table is consistent with the electronic watermark data number of taking-up;

And an electronic watermark data detector, it extracts m data from described electronic watermark data table and described extracted data table, m=1 wherein, and 2,3 ... n, and calculating is extracted the statistical approximation of data and electronic watermark data.

2. method that detects the electronic watermark data in the digital picture is characterized in that comprising following steps:

Prepare electronic watermark data in advance and extract the position table, the view data that its record will comprise n electron-like watermark data is of a size of the position of taking out in the piece of j * k pixel from image, wherein j and k are natural number, and the electronic watermark data table, its storage n electron-like watermark data;

Take out image block from digital picture, carry out orthogonal transform, and obtain the image block of taking-up and the positional information of the data after the orthogonal transform;

On the basis of described positional information, extract position table taking-up and represent the positional information that each class of the electronic watermark data of n class is taken out from image block from described electronic watermark data, take out the frequency data that comprise electronic watermark data from frequency data, and will the described frequency data that comprise electronic watermark data store into and extract the precalculated position that the position is shown with the corresponding to electronic watermark data of the electronic watermark data number that is taken out by the piece of described quadrature transformer output;

Extract the position table from described electronic watermark data table and described electronic watermark data and take out m data, m=1 wherein, 2 ... n, and the data that are extracted of calculating and the statistics similarity degree of electronic watermark data; And

Each of a described n electronic watermark data is estimated the size of this statistics similarity degree.

Images