JP5142942B2 - Digital watermark detection method, apparatus and program - Google Patents

Description

The present invention relates to a method, an apparatus, and a program for detecting an electronic watermark by inputting an image including an original image and an electronic watermark.

  As a digital watermark insertion method for inserting a digital watermark into an image, the applicant of the present application uses a first predetermined number of non-zero frequency components for detecting an enlargement / reduction ratio of an image in a frequency domain as a first circle. A non-zero frequency component for detecting the rotation angle of the image in the frequency domain, and a second predetermined number of non-zero different from the first predetermined number. Is a second circumference different from the first circumference, and the image enlargement / reduction ratio is set on the second circumference whose radius is in a predetermined proportional relationship with the radius of the first circumference. The step of generating at a position determined relative to the position of the first predetermined number of non-zero frequency components for detection is different from the first circumference and the second circumference in the frequency domain. A third circumference with a radius of the first circumference and a radius of the second circumference; Of the third predetermined number of candidate positions that are on a third circumference having a constant proportional relationship and are determined relative to the non-zero frequency component for detecting the rotation angle of the image, the third Generating a non-zero frequency component at a position corresponding to the value of the digital watermark data to be inserted by a fourth predetermined number that is less than the predetermined number and different from the first predetermined number, and the generated non-zero frequency component And a step of converting from the frequency domain to the spatial domain, and a step of adding or subtracting the image obtained by the conversion to some or all blocks of the original image. doing.

In addition, as a digital watermark detection method for detecting a digital watermark inserted in an image, the applicant of the present application divides the image into a plurality of blocks, and converts the data of each block from the spatial domain to the frequency domain. Or after the conversion, obtaining the amplitude of each frequency component of each block, adding the amplitude of each frequency component between the blocks, obtaining the total amplitude for each frequency, and frequency In the region, among the total amplitude, the total amplitude corresponding to the first predetermined number of non-zero frequency components for detecting the enlargement / reduction ratio of the image and corresponding to the non-zero frequency components positioned on the first circumference Detecting the magnification ratio of the image by detecting the first predetermined number for detecting the rotation angle of the image in the total amplitude in the frequency domain A second predetermined number of non-zero frequency components having a radius on a second circumference having a predetermined proportional relationship with the radius of the first circumference, the first predetermined number of non-zero frequency components Detecting a rotation angle of the image by detecting a total amplitude corresponding to a non-zero frequency component at a position determined relative to the position of A third predetermined number of candidate positions is a third circumference different from the first circumference and the second circumference, and the radius is the first, based on the total amplitude whose positions are corrected based on the rotation angle. On the third circumference which is in a predetermined proportional relationship with the radius of the second circle and the radius of the second circle, and is determined relative to the non-zero frequency component for detecting the rotation angle of the image. The third candidate position is detected by using the relationship that the second predetermined number of candidate positions are detected. Is inserted based on a combination of positions of a fourth predetermined number of total amplitudes that are less than the third predetermined number and show a relatively large value among them, and differ from the first predetermined number. A digital watermark detection method characterized by comprising the step of detecting the detected data (see, for example, Patent Document 1).
JP 2004-15396 A

  However, in the conventional digital watermark detection method, since the image obtained by superimposing the digital watermark on the original image is directly Fourier transformed, the frequency component obtained by the Fourier transformation includes the original image. Frequency components were also included. The frequency component of the original image becomes an obstacle, and the digital watermark before being embedded in the original image cannot be detected with high accuracy.

  In addition, when the digital watermark detection method is applied to an image photographed by a camera, the camera performs auto white balance processing, so that the detection accuracy of the digital watermark may be lowered due to the influence of the auto white balance processing. This is because the light / dark difference produced by inserting the digital watermark is reduced by the auto white balance process.

SUMMARY OF THE INVENTION An object of the present invention is to provide a digital watermark detection method, an apparatus thereof, and a program thereof for enabling detection of a digital watermark embedded in an original image with high accuracy.

In particular, the present invention provides a digital watermark detection method, apparatus and program for detecting a digital watermark embedded in an original image with high accuracy from an image subjected to auto white balance processing. The purpose is to provide.

According to the present invention, a digital watermark detection method for detecting a digital watermark that is embedded in a predetermined frequency component in the frequency domain and then subjected to inverse Fourier transform and then added or subtracted to or from the original image in the spatial domain. a is, within the above and the level of each target pixel in the spatial domain, the scope or 5 pixels × 5 pixels 3 pixels × 3 pixels from the target pixel of an image included with the original image and the watermark A difference calculating step for calculating a difference between a level obtained in accordance with a predetermined algorithm based on a level of surrounding pixels in a predetermined range, and image data obtained by collecting the difference in the frequency domain without using non-linear processing. after converting the data, the electronic watermark detecting method comprising Rukoto to have a, a detection step of detecting the electronic watermark is provided.

Further, according to the present invention, a digital watermark for detecting a digital watermark that is embedded in a predetermined frequency component in the frequency domain and then subjected to inverse Fourier transform and then added or subtracted to the original image in the spatial domain. a detection device, wherein the level of each target pixel in the spatial domain, the scope or 5 pixels × 5 pixels 3 pixels × 3 pixels from the target pixel of an image included with the original image and the watermark A difference calculating means for calculating a difference between a level obtained in accordance with a predetermined algorithm based on a level of surrounding pixels in a predetermined range, and the frequency of the image data obtained by collecting the difference without using non-linear processing There is provided a digital watermark detection apparatus comprising: detection means for detecting the digital watermark after conversion into data in a region .

Furthermore, according to the present invention, there is provided a digital watermark detection program for causing a computer to execute the above digital watermark detection method.

  According to the present invention, the influence of the original image can be eliminated as much as possible in the detection of the digital watermark. Further, according to the present invention, the influence of the auto white balance process can be eliminated as much as possible.

  The best mode for carrying out the present invention will be described below in detail with reference to the drawings.

  FIG. 1 shows some pixels of an image into which a digital watermark is inserted according to an embodiment of the present invention.

  In the present embodiment, the digital watermark included in each pixel of the image in which the digital watermark is inserted is separated from the original image. Accordingly, although the digital watermark is separated for each pixel, each pixel that is the subject of separation at that time is referred to as a target pixel. Then, pixels other than the target pixel within a predetermined range when viewed from the target pixel are referred to as peripheral pixels. In the example of FIG. 1, the predetermined range is a range of 3 vertical pixels × 3 horizontal pixels centered on the target pixel, and therefore the number of peripheral pixels is 8, but the predetermined range is centered on the target pixel. Thus, the range of 5 pixels × 5 horizontal pixels may be set, so that the number of peripheral pixels may be 24. In FIG. 1, the target pixel is marked with a circle, and the peripheral pixels are marked with a triangle. If the predetermined range is vertical 7 pixels × horizontal 7 pixels, it is difficult to separate the digital watermark from the original image. Therefore, the predetermined range is centered on the pixel of interest and is vertical 3 pixels × horizontal 3 pixels or It is preferable that the length is 5 pixels × 5 pixels.

Next, as shown in Equation (1) in FIG. 2, the blue level of the target pixel is represented as B _{i, j} . Further, as shown in Expression (2) of FIG. 2, the blue level of the peripheral pixels is represented as _{Bk, l} . The digital watermark is inserted into B (blue) of the three primary colors RGB. However, it is not always necessary to insert a digital watermark into B (blue), and a digital watermark may be inserted into R (red) or G (green).

  Further, the level refers to any value in the range of 0 to 255 for each of RGB when each of RGB of the image is expressed with 256 gradations.

[Embodiment 1]
In the first embodiment, as shown in Expression (3) of FIG. 2, a simple average of the blue levels of eight peripheral pixels in the range of 3 vertical pixels × 3 horizontal pixels centered on the target pixel is used as the average level of the peripheral pixels. _Let B _{ave i, j} .

The difference D _{i, j} for the blue level of the pixel of interest is calculated by equation (4) in FIG.

It is possible to detect a digital watermark embedded in a frequency component by Fourier-transforming the differences D _{i, j} for all pixels of interest in the imaging range.

[Embodiment 2]
In the second embodiment, various conditions are classified, and the weighted average level W _{i, j} of the peripheral pixels is obtained instead of the simple average B _{ave i, j} of the peripheral pixels. After that, the difference D _{i, j} is obtained and Fourier transformed, as in the first embodiment. However, when obtaining the difference D _{i, j} , the weighted average level W _{i, j} is used instead of the simple average B _{ave i, j} .

Next, a method for calculating the weighted average level W _{i, j} of the peripheral pixels will be described.

First, the quasi-luminance I _{i, j} of the pixel of interest is defined as shown in equation (5) in FIG. The quasi-luminance is a linear combination of an R (red) component and a G (green) component, which are primary colors, and does not include B (blue). The quasi-luminance of the peripheral pixels is represented by replacing the subscripts i and j with other subscripts.

The average values I _{ave i, j} and standard deviations I _{sd i, j} of the peripheral pixels viewed from the target pixel are defined as shown in Expression (6) and Expression (7) in FIG. 3, respectively.

The weighted average _{Wi, j} is defined by equation (8) in FIG. As is clear from the equation (8), the weighted average _{Wi, j} is obtained by adding the numerical value S to the simple average.

The numerical value S is defined by equation (9) in FIG. As is clear from equation (9), S _{k, l} (eg, S _{i−1, j−1} ) is
A condition that the standard deviation I _{sd i, j} is less than a predetermined threshold TH;
A condition that the quasi-luminance I _{i, j of the} pixel of interest is larger than I _{ave i, j} −I _{sd i, j and} smaller than I _{ave i, j} + I _{sd i, j} ,
And quasi-luminance I _k of the peripheral _pixel, quasi luminance I _i of the target pixel from _l, blue level B _k values and peripheral pixels obtained by subtracting the _j, blue level B _i of the pixel of interest from _l, by subtracting _j obtained The product of the value with the value is greater than zero,
It is 1 for the peripheral pixels where all of the above are satisfied, and zero for the other peripheral pixels.

  The value of the coefficient α appearing in the equation (9) is not particularly defined, but may be 1, for example.

Summarizing Equation (9) and Equation (10), the value added to the weighted average _{Wi, j} is divided into cases as shown in FIG.

Referring to FIG. 5, Case (1) is a case where there is a high possibility that B _{k, l} −B _{i, j} is generated due to a digital watermark.

Case (2) is a case where B _{k, l} may be affected by auto white balance by the camera.

Case (3) is a case where B _{k, l} is not significantly affected by the auto white balance by the camera, but the possibility of receiving it cannot be denied.

Case (4) is a case where B _{k, l} may be affected by the auto white balance by the camera _, but there is a possibility that the value of the digital watermark embedded intentionally can be extracted correctly.

  Next, a method for detecting a digital watermark according to the above-described concept will be described.

  FIG. 6 is a flowchart illustrating a method of detecting a digital watermark according to the second embodiment.

  Referring to FIG. 6, first, an image in which a digital watermark is inserted is captured by a camera (step S101).

  Next, the process sandwiched between steps S103S and S103E is repeated for all pixels of the captured image.

  In each repetition, the digital watermark is separated from the original image (step S105). Details of this step will be described later.

  When the repetition of the process between step S103S and step S103E is completed, the separated digital watermark is Fourier transformed (step S107).

  Next, the frequency component in which the digital watermark is embedded is detected (step S109).

  Next, digital watermark data is detected from the detected frequency component (step S111).

  FIG. 7 is a flowchart showing details of step S105 shown in FIG.

  Referring to FIG. 7, first, the quasi-luminance of the target pixel is calculated (step S105-11).

  Next, the quasi-luminance of surrounding pixels is calculated (step S105-13).

  Next, a simple average value of the quasi-luminances of the peripheral pixels is calculated (step S105-15).

  Next, the standard deviation of quasi-luminance of surrounding pixels is calculated (step S105-17).

  Next, the weighted average level of surrounding pixels is calculated (step S105-19). Details of this will be described later.

  Next, the difference between the level of the pixel of interest and the weighted average level of surrounding pixels is calculated (step S105-21).

  Next, an offset level is added to the difference calculated in step S105-21 (step S105-23). For example, the offset level may be 128 if RGB is expressed by 256 gradations.

  FIG. 8 is a flowchart showing details of step S105-19 shown in FIG.

  Referring to FIG. 8, first, the weighted average level is initialized to zero (step S105-19-1).

  Next, an unweighted average level (normal average level) is added to the weighted average level.

  Next, it is determined whether or not the standard deviation of the quasi-luminances of the peripheral pixels is less than a predetermined threshold (step S105-19-5).

  If the standard deviation of the quasi-brightness of the peripheral pixels is equal to or greater than a predetermined threshold (NO in step S105-19-5), the process ends.

  When the standard deviation of the quasi-luminances of the peripheral pixels is less than a predetermined threshold (YES in step S105-19-5), the processing sandwiched between steps S105-19-7S and step S105-19-7E for each peripheral pixel repeat.

  In each repetition, first, the quasi-brightness level of the target pixel is larger than the value obtained by subtracting the standard deviation of the peripheral pixels from the quasi-brightness average value of the peripheral pixels, and becomes the average value of the quasi-luminances of the peripheral pixels. It is determined whether or not the value is smaller than the value obtained by adding the standard deviation of the peripheral pixels (step S105-19-9).

  If the determination result in step S105-19-9 is negative, the process for the current peripheral pixel is terminated.

  If the determination result in step S105-19-9 is affirmative, the value obtained by subtracting the quasi-luminance level of the pixel of interest from the quasi-luminance level of the surrounding pixels and the blue level of the pixel of interest from the blue level of the surrounding pixels. It is determined whether or not the product of the value obtained by subtracting the level is positive (step S105-19-11).

  When the determination result in step S105-19-11 is negative, the weighted average level is obtained by multiplying a value obtained by subtracting the blue level of the target pixel from the blue level of the surrounding pixels by a predetermined coefficient. Add the values.

  The above-described method can be performed only by hardware, but can also be performed by causing a computer to read and execute a program for causing a computer to execute each step of the above-described method.

It is a figure which shows the one part pixel of the image in which the digital watermark was inserted in embodiment of this invention. It is the figure which described the formula which shows the difference about the blue level of the pixel of interest by the Embodiment 2 of this invention, the blue level of a peripheral pixel, the average level of blue of a peripheral pixel, and the blue level of a pixel of interest. It is the figure which described the formula which shows the quasi-luminance of the pixel of interest by Embodiment 2 of this invention, the average value of the quasi-luminance of a surrounding pixel, and the standard deviation of the quasi-luminance of a surrounding pixel. It is the figure which described the type | formula which shows the weighted average level by Embodiment 2 of this invention. It is a figure which shows the addition value to the weighted average for every case of the surrounding pixel by Embodiment 2 of this invention. 6 is a flowchart illustrating a method for detecting a digital watermark according to a second embodiment. It is a flowchart which shows the detail of the step which isolate | separates the electronic watermark shown in FIG. 6 from an original image. It is a flowchart which shows the detail of the step which calculates the weighted average level of the surrounding pixel shown in FIG.

Claims (13)

A digital watermark detection method for detecting a digital watermark that is embedded in a predetermined frequency component in a frequency domain, and then subjected to inverse Fourier transform and then added or subtracted to an original image in a spatial domain,
The level of each target pixel in the spatial region of the image contained between the original image and the watermark, the predetermined range is a range of range or 5 pixels × 5 pixels 3 pixels × 3 pixels from the pixel of interest A difference calculation step for calculating a difference between the level obtained according to a predetermined algorithm based on the level of a certain peripheral pixel, and
A step of detecting the digital watermark after converting the image data obtained by collecting the differences into data in the frequency domain without going through non-linear processing;
Digital watermark detection method comprising Rukoto to have a. An addition step of adding a predetermined DC level to the difference ;
Wherein the detection step, the image data comprising collected after adding a predetermined DC level on the difference from the converted data in the frequency domain without using the non-linear process, characterized that you detect the electronic watermark The digital watermark detection method according to claim 1 . Wherein the predetermined algorithm is a digital watermark detection method according to claim 1 or 2, characterized in that an algorithm for obtaining an average level of the peripheral pixels in the predetermined range. The algorithm, digital watermark detection method as claimed in claim 1 or 2, characterized in that an algorithm for obtaining a weighted average level of the peripheral pixels in the predetermined range. The algorithm for finding the weighted average level, a simple average level of the peripheral pixels in the predetermined range when viewed from the pixel of interest, a predetermined condition among the peripheral pixels in the predetermined range when viewed from the pixel of interest 5. The digital watermark detection method according to claim 4 , wherein the weighted average level is obtained by adding a level obtained by multiplying a difference between a level of a peripheral pixel to be satisfied and a level of the target pixel by a predetermined coefficient. . The digital watermark is embedded in a predetermined first primary color,
For each pixel of interest and each peripheral pixel in the predetermined range as viewed from the pixel of interest, a linear combination of the level of the second primary color and the level of the third primary color in which the digital watermark is not embedded is performed using a predetermined coefficient. Seeking steps,
For each target pixel, obtaining an average value and a standard deviation of the linear combination over the predetermined range as viewed from the target pixel;
For each pixel of interest, first determining whether the standard deviation of the linear combination over the predetermined range as viewed from the pixel of interest is less than a predetermined threshold;
The linear combination of the target pixels for which the result of the first determination is affirmative is the predetermined range viewed from the target pixel from the average value of the linear combinations over the predetermined range as viewed from the target pixel. The linear combination of the pixel of interest that is larger than the value obtained by subtracting the standard deviation of the linear combination over the range and the result of the first determination is affirmative is the predetermined pixel when viewed from the pixel of interest. A second determination as to whether or not the average value of the linear combination over the range is smaller than a value obtained by adding the standard deviation of the linear combination over the predetermined range as viewed from the target pixel; ,
A value obtained by subtracting the linear combination of the target pixels from the linear combination of the peripheral pixels, for each peripheral pixel viewed from the target pixel for which the result of the second determination is positive, A third determination as to whether a sign of a product of a value obtained by subtracting the first primary color level of the target pixel from the first primary color level is positive;
6. The digital watermark detection method according to claim 5 , wherein a peripheral pixel in which all of the results of the first to third determinations are positive is set as a peripheral pixel that satisfies the predetermined condition. An electronic watermark detection apparatus for detecting a digital watermark that is embedded in a predetermined frequency component in a frequency domain, and then subjected to inverse Fourier transform and then added or subtracted to an original image in a spatial domain,
The level of each target pixel in the spatial region of the image contained between the original image and the watermark, the predetermined range is a range of range or 5 pixels × 5 pixels 3 pixels × 3 pixels from the pixel of interest A difference calculating means for calculating a difference between the level obtained according to a predetermined algorithm based on the level of a certain peripheral pixel, and
Detection means for detecting the electronic watermark after converting the image data obtained by collecting the differences into data in the frequency domain without going through non-linear processing;
An electronic watermark detection apparatus comprising: An adding means for adding a predetermined DC level to the difference ;
It said detection means, the image data comprising collected after adding a predetermined DC level on the difference from the converted data in the frequency domain without using the non-linear process, characterized that you detect the electronic watermark The digital watermark detection apparatus according to claim 7 . The digital watermark detection apparatus according to claim 7 , wherein the predetermined algorithm is an algorithm for obtaining an average level of surrounding pixels in the predetermined range . Wherein the predetermined algorithm, the digital watermark detection apparatus according to any one of claims 7 to 9, characterized in that the algorithm for obtaining a weighted average level of the peripheral pixels in the predetermined range. The weighted average algorithm seeking is the simple average level of the peripheral pixels in the predetermined range when viewed from the pixel of interest, a predetermined condition is satisfied among the peripheral pixels in the predetermined range when viewed from the pixel of interest The digital watermark detection apparatus according to claim 10 , wherein the weighted average level is obtained by adding a level obtained by multiplying a difference between a level of a peripheral pixel and the level of the target pixel by a predetermined coefficient. The digital watermark is embedded in a predetermined first primary color,
Means for obtaining a linear combination of the level of the second primary color and the level of the third primary color in which the digital watermark is not embedded, using a predetermined coefficient for each peripheral pixel in the predetermined range as viewed from each target pixel;
Means for obtaining an average value and a standard deviation of the linear combination over the predetermined range as viewed from the target pixel for each target pixel;
Means for making a first determination as to whether or not the standard deviation of the linear combination over the predetermined range as viewed from the target pixel is less than a predetermined threshold for each target pixel;
The linear combination of the target pixels for which the result of the first determination is affirmative is the predetermined range viewed from the target pixel from the average value of the linear combinations over the predetermined range as viewed from the target pixel. The linear combination of the pixel of interest that is larger than the value obtained by subtracting the standard deviation of the linear combination over the range and the result of the first determination is affirmative is the predetermined pixel when viewed from the pixel of interest. Means for making a second determination as to whether or not the average value of the linear combination over the range is smaller than a value obtained by adding the standard deviation of the linear combination over the predetermined range as viewed from the target pixel; ,
A value obtained by subtracting the linear combination of the target pixels from the linear combination of the peripheral pixels, for each peripheral pixel viewed from the target pixel for which the result of the second determination is positive, Means for making a third determination as to whether or not a sign of a product of a value obtained by subtracting the first primary color level of the target pixel from the first primary color level is positive;
The digital watermark detection apparatus according to claim 11 , wherein a peripheral pixel in which all of the first to third determination results are affirmative is a peripheral pixel that satisfies the predetermined condition. A digital watermark detection program for causing a computer to execute the digital watermark detection method according to any one of claims 1 to 6 .

Images