JP2003032469A - Digital watermark detection processing device, digital watermark detection processing method, and program - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To provide a configuration for performing information detection by standardization processing in a digital watermark embedding and detection processing configuration for embedding and detecting bit information. SOLUTION: Bit information is acquired by detecting a difference between a plurality of digital watermarks, and detection is performed by using an average difference between detection values of a plurality of digital watermarks instead of using the detected values as statistics. Utilizing a statistic obtained by standardizing the value, the presence / absence of embedding of a digital watermark and bit information at the time of embedding are determined. Since the statistic used in the standardization process is a t statistic whose distribution is uniquely determined if the degree of freedom is determined, false positive or false ne
It becomes easy to make the probability of gative close to a desired value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for embedding or reading additional information such as copyright information and editing information in data such as images, and for example, electronic information as additional information that is difficult to recognize in an image in a normal observation state. The present invention relates to a digital watermark detection processing apparatus, a digital watermark detection processing method, and a program that perform processing for detecting a watermark (watermark: also referred to as Digital Watermarking or Data Hiding).

[0002]

2. Description of the Related Art With the progress of digital technology, digital recording / reproducing apparatuses have been popularized which are free from image quality deterioration and sound quality deterioration caused by repetitive execution of recording / reproducing processing.
Digital contents such as various images and music are recorded on DVD, C
It is becoming available for distribution and distribution through a medium such as D or a network.

Unlike analog recording / reproduction, digital recording / reproduction does not cause data deterioration even if recording / reproduction processing is repeatedly executed, and therefore the same quality as original data is maintained. The widespread use of such digital recording / reproducing technology results in a flood of illegal copies, which is a major problem from the viewpoint of copyright protection.

In order to deal with copyright infringement due to illegal duplication (copying) of digital contents, duplication control information for duplication control is added to the digital contents,
There has been proposed a configuration for preventing unauthorized copying by reading the copy control information at the time of recording / reproducing the content and executing a process according to the read control information.

There are various modes for controlling content duplication. For example, a typical method is CGMS (Copy).
Generation Management Sys
tem: Copy Generation Management
System) method. In the CGMS system, if it is an analog video signal (called CGMS-A), a specific one horizontal section within the vertical blanking period of the luminance signal,
For example, in the case of an NTSC signal, 2 bits of the 20-bit additional information superimposed on the effective video portion of the 20th horizontal section are superimposed as information for duplication control, and are also called a digital video signal (CGMS-D). 2), the 2-bit information for copy control is transmitted as additional information to be inserted and added to the digital video data.

In the case of this CGMS system, the meaning content of 2-bit information (hereinafter referred to as CGMS information) is [00].
...... Duplicate is possible [10] ...... It is possible to duplicate once (only one generation can be duplicated) [11] ... Duplicate is prohibited (absolute duplicate is prohibited).

The above-mentioned CGMS system is an example of a typical copy control system, and there are various other systems for protecting copyright of contents. For example, in digital broadcasting performed by a broadcasting station, program sequence information (S) included in a transport stream (TS) packet that constitutes digital data.
There is a method in which a digital copy control descriptor (Digital Copy Control Descriptor) is stored in (I: Service Information), and copy generation control according to the descriptor is performed when data received by a receiving device is recorded in a recording device.

However, the above control information is added as bit data to, for example, the header of the content, and it is difficult to completely eliminate the possibility of tampering with the added data. A digital watermark (watermark) is an advantageous configuration in terms of eliminating the possibility of data tampering. Watermarks are difficult to see or perceive in the playback state of normal contents (image data or audio data), and the detection and embedding of digital watermarks can only be performed by executing specific algorithms or processing by specific devices. It will be possible. By detecting a digital watermark (watermark (WM)) at the time of content processing in a receiver, a recording / reproducing device, and the like, and performing control according to the digital watermark, more reliable control is possible.

Digital watermark (watermark (WM))
The information embedded in the content by the content is not limited to the above-mentioned copy control information, but various information such as content copyright information, content processing information, content configuration information, content processing information, content editing information, or content reproduction processing method. Can be embedded. For example, the editing information is embedded during the content editing process, and at each editing step, the processing step is confirmed by referring to the digital watermark. Such edit information is embedded in the content as a new digital watermark at each edit step of the content, and is finally removed from the content.

[Digital watermark embedding] Various methods have been proposed for embedding and detecting digital watermarks in data. The digital watermark applied here is a method based on the statistical property of data as an original signal, for example, an image, and a digital watermark embedding process based on the statistical property of this image will be described. Let P be the original image that is the target of the electronic watermark embedding, and let W be the electronic watermark pattern embedded in the original image P. At this time,
The digital watermark pattern W is assumed to satisfy the following equation.

[0011]

[Equation 1]

As an example, the original image P and the digital watermark pattern W are set as in the following equation. That is,

[0013]

[Equation 2]

However, in the above equation, for simplicity, the original image P
Has a size of 5 × 4 pixels. Since adjacent pixels in an image generally have a value close to each other, adjacent elements of the original image P are set as close values. In the example shown in the above [Formula 2], the sizes of the original image P and the digital watermark pattern W are set to be the same, but the sizes of the original image P and the digital watermark pattern W do not necessarily have to be the same size. . If the sizes are not the same, the calculation is performed on the overlapping portion of the original image and the digital watermark pattern.

The digital watermark embedding process is executed based on the following equation.

[0016]

[Equation 3] M = P + W

Here, M represents an image in which the digital watermark pattern W is embedded in the original image P. The value of M is calculated as follows in the example shown in [Equation 2] above.

[0018]

[Equation 4]

[Digital watermark detection] This digital watermark pattern W is used for digital watermark detection. The detection of the digital watermark in the original image P in which the digital watermark W is not embedded is defined by the following equation.

[0020]

[Formula 5] x = P · W

It is defined as Here, the operator “·” is the inner product of the matrix, and x is the inner product value of the original image P and the digital watermark pattern W.

Since the total sum of the elements of the digital watermark pattern is 0 (see [Equation 1]) and the adjacent pixels of the image tend to have generally close values, the inner product value x is near 0. It becomes a value. In the example shown in [Equation 2] above,
The inner product value is as follows.

[0023]

[Equation 6]

Next, the same operation is performed on the image M in which the digital watermark is embedded. In the detection of the digital watermark for the image M in which the digital watermark W is embedded, the inner product value x ′ is obtained according to the following equation, as in the above.

[0025]

[Equation 7]

While the inner product value of the original image P and the digital watermark pattern W becomes a value near 0, the inner product value x'of the image M in which the digital watermark is embedded and the digital watermark pattern W is
It is close to the inner product value of the digital watermark pattern W itself. That is,

[0027]

[Equation 8] WW

It becomes a value near the above formula [Equation 8]. This inner product value W · W can be used as a measure of the embedding strength of a digital watermark. When the target inner product value W · W when embedding the digital watermark pattern is large, the digital watermark embedding strength is expressed as “strong”, and when the inner product value W · W is small, the digital watermark embedding strength is expressed as “weak”. To do.

Further, when the absolute value of the inner product value x of the original image P and the digital watermark pattern W and the inner product value x ′ of the image M in which the digital watermark is embedded and the digital watermark pattern W becomes a large value, the digital watermark is detected. The strength is expressed as “strong”, and the detection strength of the digital watermark is expressed as “weak” when the absolute value of the inner product value x or the inner product value x ′ becomes a small value.

Further, the correlation between the image and the digital watermark pattern is "large" or "high" when the detection strength of the digital watermark is strong, and the correlation between the image and the digital watermark pattern is "low" when the detection strength of the digital watermark is weak. It may be expressed as "small" or "low".

When the inner product value x of the original image P and the digital watermark pattern W and the inner product value x ′ of the image M in which the digital watermark is embedded and the digital watermark pattern W are obtained in various images,
The relative frequency distributions thereof are represented by the probability density functions f and f ′ and are as shown in FIG.

In determining whether or not a digital watermark is embedded in an image, the inner product value x of the image P in which the digital watermark is not embedded and the digital watermark pattern W is distributed around 0, and It is used that the inner product value x ′ of the image M in which is embedded and the digital watermark pattern W is distributed around W · W which is the inner product value of the digital watermark pattern W itself. The inner product value x ″ of the image to be checked for the presence or absence of the digital watermark and the digital watermark pattern W is obtained, and the presence or absence of the digital watermark pattern is determined by comparing with a certain threshold value (th). Specifically, the following formula can be applied. Is.

[0033]

[Equation 9] x "<th then no-watermark x ”≧ th then watermarked

In the above equation, the inner product value x ″ of the image for which the presence or absence of the digital watermark is to be checked and the digital watermark pattern W is the threshold value (t
When it is smaller than h), no watermark pattern is embedded. When the inner product value x ″ is equal to or larger than the threshold value (th), it is determined that the digital watermark pattern is embedded, which is illustrated in FIG.

This threshold value (th) is a value determined from the statistical properties of the probability density function f of the inner product value x and the probability density function f'of the inner product value x '. It is called "false positive" that the threshold value (th) is exceeded even if the digital watermark is not embedded, that is, it is determined that the digital watermark is embedded even though the digital watermark is not embedded. On the contrary, it is called “false negative” that the threshold is not satisfied even though the digital watermark is embedded, that is, it is determined that the digital watermark is not embedded even though the digital watermark is embedded. When the probability p _{FP of} "false positive" and the probability p _FN of "false negative" which are such erroneous determinations are determined, the threshold value (t
h) is fixed.

In an application that uses a digital watermark
The required probability p_FPAnd probability p_FNThe threshold th that satisfies
Problems can arise in defining. Detection value WW first
, The probability p_FPThreshold trying to lower
If th is increased, the probability p_FN, And conversely, the probability p_FN
If the threshold th is decreased in an attempt to decrease_FPIs above
There is a relationship of being angry. Probability p_FPAnd probability p_FNIs a trade
They are off and cannot be lowered at the same time. Figure 3
, The probability p is shifted by shifting the threshold th to the right._FPGoes down,
Probability p_FNGoes up. If the threshold th is shifted to the left, the probability p_FNBut
Down, probability p _FPGoes up.

When determining the threshold value th, it is necessary to adjust the detection value W · W at the time of embedding so that the threshold value th satisfying the required probability p _FP and probability p _FN can be set.
That is, as shown in FIG. 4, it is necessary to be able to adjust the center position of the probability density function f ′ representing the detection value in the image in which the digital watermark is embedded, at the time of embedding. When the probability p _FP and the probability p _FN are larger than the required values, the center position is shifted to the right,
When the probability p _FP and the probability p _FN are smaller than the required values, the center position is shifted to the left.

In order to enable such adjustment, the embedding process of the digital watermark pattern W in the original image P is changed from the above equation [Equation 3] to the following equation. That is,

[0039]

[Equation 10] M = P + cW

Change to In the above equation, c is a non-negative scalar value. Due to this change, the embedding strength of the digital watermark is set to the inner product value W · W of the digital watermark pattern W itself.
To a value obtained by multiplying by a scalar value, that is, the following expression.

[0041]

[Equation 11] cW ・ W

In order to determine the threshold, first, the probability p _FP and the probability p _FN required in the application using the digital watermark are determined, and the boundary value th _FP and the boundary value th _{FN at} that time are determined.
Determine. Depending on the application using the digital watermark, an extremely small value may be required for the probability p _FP and the probability p _FN . Especially, the probability p _FP is significantly required. For example, when using a digital watermark for copyright protection,
This is because even if the illegally copied image is mistakenly determined to be a legal image, it is unlikely to lead to a user complaint, whereas if the wrongly determined illegally duplicated image is an illegally duplicated image, it is likely to lead to a user complaint. .

Ideally, the probability p _FP and the probability p _FN are as close to 0 as possible. However, the threshold th
, The strength cW · W of the digital watermark that must be embedded increases, and the effect of the digital watermark embedding on the image quality cannot be ignored. There is a trade-off between the reliability of digital watermark detection and the effect of digital watermark on image quality.

While ensuring the reliability of digital watermark detection,
Various methods have been proposed to suppress the influence of digital watermarks on the image quality as much as possible. As a widely used basic method, there is a method of embedding strongly in the edge part of the original image and weakly embedding it in the flat part. When the change of the pixel value is numerically the same, the change is easily noticeable in the flat part, but the change is hard to be noticeable in the edge part, which utilizes the human visual characteristic. Even if the embedding strength of the digital watermark is set to be different between the edge part and the flat part, if the digital watermark pattern with a certain strength is embedded in the entire image, the reliability of the detection of the digital watermark is almost the same. Becomes

[Multi-bit embedding information] The contents up to this point are a method of embedding a digital watermark pattern in an image and a method of detecting the digital watermark. With this digital watermark embedding detection method, it is possible to determine only two ways, whether the digital watermark pattern is embedded in the image or not. In other words, only 1-bit information can be expressed. In the following, we will touch on how to make the embedded information multi-bit.

The method of embedding multi-bit information in an image with a digital watermark is roughly classified into a method of using a plurality of digital watermark patterns, a method of dividing an image into small regions, and a combination of these.

In the method using a plurality of digital watermark patterns, a method of expressing desired information by giving different meanings to each of a plurality of digital watermark patterns and exclusively embedding it in an image, and a method of simultaneously superposing a plurality of digital watermark patterns. A method of embedding desired information by embedding it in an image and combining it, and a method of combining these two methods are considered. FIG. 5 shows how a plurality of digital watermark patterns are embedded in the original image.

A method of expressing desired information by embedding a plurality of digital watermark patterns in different meanings in each image and embedding it exclusively in an image is necessary when the number of bits of the information to be embedded in the image is b. The type n of the digital watermark pattern is n = 2 ^b . On the other hand, in the method of simultaneously embedding a plurality of digital watermark patterns and embedding them in an image to express desired information, the required digital watermark pattern type n is n = b. However,
Although the latter requires a small number of types of digital watermark patterns, it is often necessary to take appropriate measures against image deterioration because multiple digital watermark patterns are embedded in an image. Finally, in the method in which these two methods are combined, the required digital watermark pattern type n is b ≦ n ≦ 2 ^b , which has the characteristics of both methods.

The method of dividing an image into small areas is another method of embedding multi-bit information in an image with a digital watermark. By giving different functions to each small area, a plurality of electronic images can be included in the image. The watermark is intended to exist at the same time. Various methods for arranging the small areas have been proposed. Here, an example in which the small regions are arranged in a grid pattern as shown in FIG. 6 will be described. I and j in FIG. 6 are non-negative integers.

When dividing an image into small areas, the number of divisions becomes a problem. When the information to be embedded in the image is b bits, a method of dividing the image into b small areas can be considered first. However, when embedding a digital watermark pattern in various images, the visual characteristics of the image are considered. Since this method often embeds a digital watermark pattern, this method is prone to problems. For example, when adding processing such as embedding strongly in the edge part of the embedded image and weakly embedding it in the flat part of the image,
If the small area corresponding to a certain bit happens to be a flat area, the digital watermark embedded in that area may not be detected. Even if there is a region in which digital watermark detection fails even in one region, even if the digital watermark embedded in the remaining region is detected, the meaning of the entire combination is lost.
When dividing an image into small regions, dividing into smaller regions than b has the advantage that the digital watermark can be detected stably in various images. Even if the embedding strength of the digital watermark pattern becomes extremely weak in one small area, if the necessary amount of the digital watermark pattern is secured in the remaining small area in which the same bit information is embedded, the detection of the digital watermark as a whole is performed. Is possible.

FIG. 7 shows an example of division of a small area when the embedded information is 8 bits. A plurality of small areas corresponding to the same bit are allocated in the image. In the figure, b
_k (k = 0, 1, 2, ..., 7) is a small area for embedding a digital watermark pattern that represents bit information of the k-th bit.

[Expression of Bit Information] By the above-mentioned method, it is possible to express bit information depending on the presence or absence of a digital watermark. However, if bit information is expressed depending on the presence or absence of a digital watermark, inconvenience may occur. For example, if it is assumed that bit information = 1 when a digital watermark is detected and bit information = 0 when it is not detected, then bit information = 0 as defined when a digital watermark is not detected at the time of detection. It may not be possible to distinguish between the original image and the original image.

As a method for avoiding such a situation, a method of expressing bit information as shown in FIGS. 8 and 9 is known. In FIG. 8, the detection value of the digital watermark is to be distinguished based on its code, and in FIG. 9, two digital watermark patterns are prepared and to be distinguished by the combination of the detection results. When two digital watermark patterns are used, they can be embedded in the same region in an overlapping manner or separately in different regions.

In the method for distinguishing the detected value of the digital watermark shown in FIG. 8 based on its code, when bit information = 1, [Equation 10], that is, M = P + cW, is used as it is, but bit information = 0. In this case, the digital watermark pattern embedding method is changed from [Equation 10] to the following equation.

[0055]

(12) M = P-cW

In this way, the method of embedding the digital watermark pattern is changed according to bit = 0,1. On the other hand, there is no change in the digital watermark detection method, and if the detected value is within the range of the threshold value-th to th, it is not detected and the detected value is the threshold value th.
In the above case, it is determined that bit information = 1 is detected, and when the detected value is equal to or less than the threshold value-th, it is determined that bit information = 0 is detected. In other words, digital watermark bit information 0,
The determination process 1 and the determination that the electronic watermark is not detected are processes based on the following equations.

[0057]

[Equation 13] Is.

In the method of FIG. 9, the embedding method of each digital watermark pattern is [Equation 10], that is, M = P.
+ CW remains unchanged, but two digital watermark patterns 1 and 2 are used to represent 1 bit. For example, when bit information = 1, the digital watermark pattern 1 is embedded and the digital watermark pattern 2 is not embedded as shown in FIG. When bit information = 0, the digital watermark pattern 1 is not embedded as in FIG. 9B, but the digital watermark pattern 2 is embedded. At the time of detection, the two digital watermark patterns are independently detected, and the bit information is determined from the respective detection results. That is, the following determination process is executed.

The detection value of the digital watermark pattern 1 is the threshold th.
_{When it is 1} or more and the detected value of the digital watermark pattern 2 is less than the threshold value th ₂ , the detected bit information = 1. When the detected value of the digital watermark pattern ₁ is below the threshold th ₁ and the detected value of the digital watermark pattern 2 is equal to or larger than the threshold th ₂ , the detection bit information = 0. When both the detection value of the digital watermark pattern 1 and the detection value of the digital watermark pattern 2 are below the threshold value, the bit information is not detected. This is the case when the detection process is performed on the original image.

If both the detection value of the digital watermark pattern 1, the detection value of the digital watermark pattern 2, and the detection values of both of the two digital watermark patterns are equal to or more than the threshold value, it is determined that the value is an abnormal value, and in this case, it is not detected. And The following is a summary of these determination processes.

[0061]

X ₁ ≧ th ₁ and x ₂ <th ₂ then 1 x ₁ <th ₁ and x ₂ ≧ th ₂ then 0 Otherwise then no watermark detected

According to the above-described digital watermark bit determination method shown in FIGS. 8 and 9, both determine whether or not a digital watermark is embedded, and if embedded, the embedded bit information is 1 It is possible to determine whether it is 0 or 0.

[Flow of embedding digital watermark pattern]
FIG. 10 shows a processing example in the apparatus that executes the processing of embedding a digital watermark pattern in an image. The digital watermark pattern generation unit 1004 uses the embedded information 100 embedded in the image.
2 and digital watermark pattern generation key (key) storage unit 1003
A digital watermark pattern is generated from the digital watermark pattern generation key (key).

The embedded information 1002 is information embedded as a digital watermark, and is arbitrary information such as copy control information, copyright information, and edit information. Specifically, the digital watermark pattern generation key (key) corresponds to the method of dividing the image into small areas described in FIG. 6 when embedding the digital watermark pattern in the image, and the bits of the embedded information described in FIG. 7. Image division information, bit array information, etc., such as how to allocate small areas to be embedded, and processing information necessary to generate the embedded information 1002 as a digital watermark pattern.

The digital watermark pattern embedding unit 1005 embeds the digital watermark pattern generated by the digital watermark pattern generating unit 1004 in the original image 1001. It is this digital watermark pattern embedding unit 1005 that adjusts the embedding strength of the digital watermark pattern in the edge portion or flat portion of the original image. The image in which the digital watermark pattern is embedded is output as a digital watermark embedded image 1006.

[Flow of Digital Watermark Detection] FIG. 11 shows a processing example in the apparatus for executing the digital watermark detection processing. The digital watermark pattern generation unit 1103 generates a digital watermark pattern from the digital watermark pattern generation key (key) in the digital watermark pattern generation key (key) storage unit 1102.

The digital watermark pattern generation key (key) is
Specifically, it is image division information when embedding a digital watermark pattern in an image, bit array information, and the like, and is information necessary for detecting a digital watermark pattern.

The detection unit 1104 detects the digital watermark of the input image 1101 using the digital watermark pattern generated by the digital watermark pattern generation unit 1103. The information detected by the detection unit 1104 is output as detection information 1105.

[Bit Information Judgment Processing] With reference to FIGS. 8 and 9, it is judged whether a digital watermark is embedded and, if embedded, the bit information embedded is either 1 or 0. The bit determination method using a digital watermark that can determine whether or not has been described. In the conventional method using two digital watermark patterns, two digital watermarks are detected independently, each detected value is independently compared with a threshold value, and the embedded bit information is determined by the combination of the results. To do. The determination process is summarized as a flow shown in FIG. The flow of FIG. 12 will be described.

First, the detection values of the two digital watermark patterns are obtained independently (S1201, S1).
203). Then, it is compared with a threshold value set for each of the two digital watermark patterns to determine whether the threshold value is equal to or more than the threshold value or not (S1202, S1204).

Finally, the presence or absence of a digital watermark and the embedded bit information are determined by the combination of the determination results obtained for each of the two digital watermark patterns (S
1205). FIG. 12 illustrates the processing in the detection unit 1104 of FIG.

[Standardization of Detection Value] The detection process of the digital watermark pattern is executed in a plurality of image regions in one image, as shown in FIGS. 5, 6 and 7, and further, A moving image may be continuously executed in a plurality of consecutive image frames. That is, many detection values are acquired for a certain digital watermark pattern. A more stable bit determination process can be performed by comprehensively determining a large number of these detection values. One of the methods is standardization processing.

For example, it is assumed that the probability density function f indicating the relative frequency distribution of the detection value x of the digital watermark pattern W in the original image P is as shown in FIG. 13C when evaluated for various images. Here, even if the probability density function f for various types of images is shown as in FIG. 13C, FIG. 13 is used for continuous images within a limited period.
There is a problem of showing different forms as in (a) and (b).

For example, in the case of a movie on the theme of space, it can be expected that the space is projected for a long period of time. Similarly, for a movie on the theme of forest, a movie on the theme of forest in the space Then many seas will appear. In this way, if there is a certain theme such as a movie or TV program, there is a high possibility that similar scenes will be used repeatedly, and as a result, the probability density function obtained for a wide variety of images will be obtained. There is a possibility that the relative frequency distribution obtained within the specific detection period as a result of the electronic watermark detection process for a specific image of f is significantly different.
Ideally, it is preferable to apply a statistic that gives stable results to various images.

Generally, when the random variable X according to the average value μ and the variance σ ² is considered, the average value [X] is expressed by the following equation.

[0076]

[Equation 15]

The central limit theorem shows that the average value [X] has a normal distribution N (μ, σ ² / n) if n is sufficiently large. Here, n is a natural number and N (μ, σ ² /
n) represents a normal distribution with mean μ and variance σ ² / n. However, if the distribution of the random variable X is a normal distribution, then the distribution of the average value [X] is exactly a normal distribution. The standardized value Z of this average value [X] is calculated by the following formula.

[0078]

[Equation 16]

The standardized value Z has a standard normal distribution N (0,1). However, when it is applied to detection of a digital watermark, it cannot be applied unconditionally as it is because the population variance σ ² is unknown. An example of a method for estimating population variance will be shown. The sample variance s ² of the detected value x of the image and the digital watermark pattern is defined by the following equation.

[0080]

[Equation 17]

In the case of substituting the population variance σ ² by the above equation, the equation [Equation 16] of the standardized value Z of the average value [X] is given.
Will be given by:

[0082]

[Equation 18]

The above equation is a t statistic, and its distribution is a t distribution with n-1 degrees of freedom. The t distribution is a distribution very similar to the standard normal distribution, and like the standard normal distribution, it has a mean value of 0 and is bilaterally symmetric. When n → ∞, the t distribution matches the standard normal distribution.

Instead of using the distribution of the detected values x, the distribution of the t statistic will be used as the distribution used as a criterion for determining whether or not the digital watermark pattern is embedded. In order to obtain one t statistic in one image frame, n detection values x are required according to the definition of [Equation 15]. Therefore, one frame of the image is divided into small areas as shown in FIG.

In FIG. 14, the division is performed in a grid pattern, but a division method other than the grid pattern may be used. The same digital watermark pattern is embedded in each small area. At the time of detection, the average value [x] of the detection values x in small area units and the variance s ² are obtained. The population mean μ of the distribution of the detected values x with respect to the original image is μ = from the definition of the watermark.
Since it is 0, the t statistic is calculated by the following equation.

[0086]

[Formula 19]

Since the t distribution is uniquely determined by the degree of freedom n−1, the t distribution obtained can be uniquely determined by determining the number n of the small areas. From the above, it can be said that the t distribution used for determining the detection of the digital watermark is stable for any original image. However, as is clear from the definition,
The variance s ² is not allowed to be s ² = 0. For this reason,
Exception handling is required when variance s ² = 0.

The reason for using the t statistic is to avoid fluctuations in the dispersion of the detection values of the digital watermark pattern and the image, and to average the detection values from the original image P and the digital watermark pattern embedded image M.
This is because it is possible to uniformly compare the differences in the average of the detected values from. For example, if you want to know whether a digital watermark is embedded or not, use the null hypothesis μ
= 0, the alternative hypothesis is set to μ ≠ 0, and the test is performed under a certain significance level. The significance level here is equal to the probability p _FP of false positive. When the null hypothesis is rejected, that is, when the absolute value of the t statistic is greater than or equal to the threshold value, μ ≠ 0
It is highly probable that the digital watermark is embedded. When the null hypothesis is not rejected, that is, t
When the absolute value of the statistic is less than the threshold value, the probability of μ = 0 is high, and it can be determined that the digital watermark is not embedded. This procedure is commonly called the t-test. Since the distribution of t is uniquely determined when the degree of freedom is determined, there is an advantage that the probability of false positive or false negative can be easily controlled.

FIG. 15 shows an example in which standardization is required. Figure 15
Shows the case where there is a difference in the distribution of the detected values of the image and the digital watermark pattern depending on the image. The distribution 1201 is the detection value for the original image, and the distribution 1202 is the detection value for the image in which the digital watermark is embedded.
Similarly, the distribution 1203 is the detected value for the original image,
A distribution 1204 is a detection value for an image in which a digital watermark is embedded. Distribution 1201 and distribution 1202
Has a large variance, and distributions 1203 and 1204 have a small variance.

It is assumed that the distributions 1202 and 1204 of the detected values of the image in which the digital watermark is embedded have the same average value. At this time, the conventional methods that do not use standardization have treated both as equivalent. However, in reality, it is expected that the probability p _FP and the probability p _FN in the distribution 1201 and the distribution 1202 are larger than desired values, and the distribution 12
03 and the distribution 1204, it is expected that the probability p _FP and the probability p _FN are smaller than desired values. In other words, it is required to determine the presence / absence of a digital watermark in consideration of not only the average value but also the variance.

FIG. 16 shows an example in which the distribution example shown in FIG. 15 is standardized. The distributions 1301 to 1304 are reprints of the distributions 1201 to 1204 in FIG. Distribution 1305 and Distribution 1306 are Distribution 1301 and Distribution 1302
Is a standardized distribution with respect to the distribution 1301. Similarly, the distributions 1307 and 1308 are distributions obtained by standardizing the distributions 1303 and 1304 with respect to the distribution 1303. It can be seen that the standardization unifies the dispersion and the presence or absence of the digital watermark can be determined by the threshold value th _Z. or,
The standardization of the distribution 1302 having a large variance is a distribution 1306 having a small average value, and the standardization of the distribution 1304 having a small variance is a distribution 1308 having a large average value.

As can be seen from FIG. 16, when the threshold value th _Z for standardization is defined by the probability p _FP , the probability p _FN changes due to the variance of the original distribution. Distribution 1306
_{Has a} large probability p _FN , and the distribution 1308 has a small probability p _FN . The probability p _FN is large like the distribution 1306, and if
When the probability p _FN becomes a value outside the allowable range, it is determined that the digital watermark has not been detected.

However, as is clear from the above [Equation 16], the variance of the average [X] of the detected values becomes smaller as the value of n becomes larger. Therefore, when the digital watermark is embedded, , The average value of the standardized value Z becomes large. Therefore, even if the probability p _FN becomes a value outside the allowable range, if the frames are continuous as in a moving image, the detection values obtained from other frames are added to the calculation. The digital watermark can be detected by re-evaluating.

As a method for preparing n samples, other than the method of dividing one frame of an image into a plurality of small regions,
As shown in FIG. 7, a method of using a plurality of image frames as a set can be considered. In this case, one detection value x is obtained for each image frame, and a plurality of image frames are collectively standardized. A method that combines these two methods, that is, one image frame is divided into a plurality of small areas,
Moreover, a method of using a plurality of image frames as a set is also conceivable.

[0095]

The electronic watermark for judging whether the digital watermark described above is embedded or not, and if embedded, judges whether the embedded bit information is 1 or 0. The bit determination process using the watermark has the following problems. The problem will be described with reference to FIG. 18A and 18B,
In this example, bit information = 0 is embedded and then detected.

Before embedding a digital watermark in a certain original image, detection processing of the digital watermark 1 and the digital watermark 2 is executed. The correlation between the digital watermark 1 and the digital watermark 2 is also detected in the original image itself before the digital watermark is embedded, and the correlation may be strong or weak.

It is assumed that the correlation detection of the original image, that is, the detection value of the digital watermark 1 is as shown by the upper arrow in FIG. 18A, and the detection value of the digital watermark 2 is as shown by the solid arrow in the lower row of FIG. 18A. That is, it is assumed that the detection value obtained when the detection process is performed on the original image is x ₁ ′ for the digital watermark 1 and x ₂ ′ for the digital watermark 2.

The digital watermark pattern 2 is embedded in the original image with the intensity c ₂ W ₂ · W ₂ . That is, in FIG.
(A) The process of embedding the digital watermark pattern 2 is executed as indicated by the dotted arrow in the lower row.

An image in which this digital watermark pattern is embedded
Detecting digital watermark 1 and digital watermark 2 from the image
It The detected value is the detected value x₁, X₂And Then electronic
Watermark 1 detection value x₁Embedding the digital watermark 1
Since it is not performed, it is the same as the detected value in the original image.
X₁= X₁’ On the other hand, the detection value x of the digital watermark 2
₂Is the point at the end of the dotted arrow in the lower part of FIG.
Chi, x₂= X₂’+ C₂W ₂・ W₂Becomes

As a result, as shown in FIG. 18A, the detection value x ₁ = x ₁ 'of the digital watermark ₁ is smaller than the threshold value th ₁ , and the detection value x ₂ = x ₂ ' + c _{2 of the} digital watermark 2 is obtained. W ₂ · W ₂ becomes the threshold value th ₂ or more. Regarding this result,
4] is applied, it corresponds to x ₁ <th ₁ and x ₂ ≧ th ₂ then 0. Therefore, this detection result can be determined to be digital watermark embedding with bit information = 0.

However, in the case where the digital watermark embedded image is coded by, for example, image data communication, and the coded data is decoded, the digital watermark detection process is executed. There are also many. When the image processing is performed on the image in which the digital watermark is embedded in this way, a slight change may occur in the image data, and this change in the image data may affect the detection value of the digital watermark.

FIG. 18B shows a case where a certain original image is shown in FIG.
8 (a) is added. In both (a) and (b), the fact that the digital watermark 2 is embedded with the strength c ₂ W ₂ · W ₂ and the bit information to be detected = 0 are the same, but the result has Reliability is very different.

In FIG. 18B, the detected value x ₁ = x ₁ ′ of the digital watermark ₁ is also the detected value x ₂ = x ₂ ′ + c ₂ W _{2 of the} digital watermark _2.
· W ₂ also, respective threshold th _1, th ₂ far from the image encoding or the like after the embedding of the digital watermark, even the threshold t if some image processing is a change in the image data is added is generated
It is unlikely that a data change exceeding h ₁ and th ₂ will occur, and thus the possibility of erroneous determination due to image processing is low.

On the other hand, in the example shown in FIG. 18A, the detected value x ₁ = x ₁ 'of the digital watermark ₁ is also the digital watermark 2
Detected value x ₂ = x ₂ '+ c ₂ W ₂ · W ₂ also has thresholds th ₁ and th ₂
When image processing is performed in the vicinity of
There is a possibility that data changes exceeding h ₁ and th ₂ may occur, and the possibility of erroneous determination of digital watermark detection by image processing increases.

As processing for preventing erroneous determination of digital watermark detection, detection processing is performed on the original image in advance when embedding a digital watermark, and the magnitude of the detected value is used to determine the embedding strength. The present applicant has proposed a configuration in which the strength of detection of an embedded digital watermark is concentrated on a value separated from a threshold by a certain value by adjusting the strength.
1-37909. By using this method, it is possible to control the embedding strength of the digital watermark pattern 2 and keep the detection value of the digital watermark 2 substantially constant, even in the images shown in FIGS. 18A and 18B, for example. Therefore, the digital watermark 2 can be detected stably without depending on the original image.

However, no improvement can be expected for the digital watermark pattern 1 in which such embedding strength adjustment processing is not performed. This is because this method is performed when embedding a digital watermark pattern. Therefore, the strength adjustment process is not executed for the digital watermark pattern 1 in which no embedding is performed, and the possibility of erroneous determination when the digital watermark is detected remains.

Further, for example, when the bit information expressing method shown in FIG.
By performing the standardization processing as shown in 6], stable detection of the digital watermark can be realized. However, in the case of the bit information expression method shown in FIG.
It is not possible to directly apply the standardization process as shown in. This is because the bit information expression method shown in FIG. 9 uses two digital watermark patterns, that is, two populations in combination, whereas the standardization process as shown in [Equation 16] is the only mother data. This is because the group is assumed.

When the standardization is not performed, the same problem as the reliability of detection due to the difference in distribution as shown in FIG. 15 in the case of the bit information expression method shown in FIG.
This may also occur in the case of the bit information expression method shown in FIG.
Examples are shown in FIGS. 19 and 20.

FIG. 19A shows the case where both the variance of the detection value of the digital watermark pattern 1 and the variance of the detection value of the digital watermark pattern 2 are both small. In this case, probability p _FP and probability p _FN
Is likely to be smaller than the desired value.
Therefore, it is highly possible that a reliable detection result is obtained.

FIG. 19B shows a case where both the variance of the detection value of the digital watermark pattern 1 and the variance of the detection value of the digital watermark pattern 2 are large. In this case, probability p _FP and probability p _FN
Is likely to be larger than the desired value.
Therefore, it is relatively likely that the detection result lacks reliability.

FIGS. 20 (a) and 20 (b) show the case where the variance of one detected value is large and the variance of the other detected value is small. Also in this case, as in the case of FIG. 19B, there is a relatively high possibility that the detection result lacks reliability. In the case of FIG. 20A, the probability p _FN is high for the digital watermark pattern 1, and in the case of FIG. 20B, the probability p _FP is high for the digital watermark pattern 2. That is, the reliability of the detection result cannot be expected.

The present invention has been made in view of the above-mentioned problems, and in a digital watermark embedding process and a detection process execution configuration for expressing bit information by a plurality of different digital watermark pattern embedding processes, digital watermark detection A digital watermark detection processing device that reduces the possibility of erroneous determination at the time, and that can perform stable digital watermark detection on various images by performing standardization processing, and a digital watermark detection processing It is intended to provide a method and a program.

More specifically, according to the present invention, a standardization process is performed in a configuration for executing digital watermark embedding and detection for expressing bit information by embedding a plurality of digital watermark patterns as shown in FIG. 9, for example. It is an object of the present invention to provide a digital watermark detection processing device, a digital watermark detection processing method, and a program capable of ensuring reliability at the time of detection for various images by performing the above.

[0114]

The first aspect of the present invention is as follows.
A digital watermark detection processing device for executing a digital watermark pattern detection process, which performs a detection process for each of a plurality of digital watermark patterns on an input image that is a target of a digital watermark pattern detection process. Digital watermark pattern detection value acquisition means for acquiring a plurality of detection values, and standardization processing of detection value differences of each digital watermark pattern based on a plurality of detection values of each pattern acquired by the digital watermark pattern detection value acquisition means Run
A standardization processing unit that acquires a standardized value calculated as a result of the standardization process, and a digital watermark embedding pattern determination unit that acquires embedded digital watermark information based on a comparison process between the standardized value acquired by the standardization processing unit and a threshold value. And a digital watermark detection processing device.

Further, in one embodiment of the digital watermark detection processing device of the present invention, the standardization processing means compares the variances based on a plurality of detection values of each pattern acquired by the digital watermark pattern detection value acquisition means. The digital watermark embedding pattern determination unit includes a variance comparison unit and a t statistic calculation unit that calculates a t statistic amount as a standardized value based on the output of the variance comparison unit, It is characterized in that the embedded digital watermark information is acquired on the basis of the comparison processing of.

Furthermore, in one embodiment of the digital watermark detection processing device of the present invention, the standardization processing means compares the variances based on a plurality of detection values of each pattern acquired by the digital watermark pattern detection value acquisition means. A variance comparison unit and a t statistic calculation unit that calculates a t statistic and a t distribution based on the output of the variance comparison unit are included, and the digital watermark embedding pattern determination unit includes the t statistic and the t distribution. The embedded digital watermark information is acquired based on a comparison process with a threshold value determined based on the above.

Further, in one embodiment of the digital watermark detection processing apparatus of the present invention, the standardization processing means applies a predetermined t statistic calculation formula to calculate the t statistic based on the output of the variance comparison unit. It is characterized in that the calculation processing is executed.

Further, in one embodiment of the digital watermark detection processing apparatus of the present invention, the standardization processing means is adapted to a case where a plurality of variances based on detection values of a plurality of digital watermark patterns are equal and a case where the variances are different from each other. The present invention is characterized in that different t statistic calculation formulas are applied to execute t statistic calculation processing based on the output of the variance comparison unit.

Further, a second aspect of the present invention is a digital watermark detection processing method for executing digital watermark pattern detection processing, wherein each of a plurality of digital watermark patterns is applied to an input image which is a digital watermark pattern detection processing target. The detection process is performed to acquire a plurality of detection values for each digital watermark pattern, and a digital watermark pattern detection value acquisition step, and a plurality of detection values for each pattern acquired in the digital watermark pattern detection value acquisition step Based on the standardization process step of performing a standardization process of the detection value difference of each digital watermark pattern, and obtaining a standardization value calculated as a result of the standardization process, a standardization value and a threshold value obtained in the standardization process step. Digital watermark embedding pattern determination that acquires embedded digital watermark information based on comparison processing In digital watermark detection processing method comprising: the step, the.

Further, in one embodiment of the digital watermark detection processing method of the present invention, the standardization processing step compares variances based on a plurality of detection values of each pattern acquired in the digital watermark pattern detection value acquisition step. A variance comparison step and a t statistic calculation step of calculating a t statistic as a standardized value based on the output in the variance comparison step, and the digital watermark embedding pattern determination step includes the t statistic and a threshold value. The embedded digital watermark information is acquired based on the comparison process of 1.

Further, in one embodiment of the digital watermark detection processing method of the present invention, the standardization processing step compares variances based on a plurality of detection values of each pattern acquired in the digital watermark pattern detection value acquisition step. A variance comparison step, and a t statistic calculation step of calculating a t statistic and a t distribution based on the output in the variance comparison step, and the digital watermark embedding pattern determination step includes the t statistic and the t distribution. The embedded digital watermark information is acquired based on a comparison process with a threshold value determined based on the above.

Furthermore, in one embodiment of the digital watermark detection processing method of the present invention, the standardization processing step applies a predetermined t-statistic calculation formula to calculate a t-statistic based on the output of the variance comparison unit. A feature is that a calculation process is executed.

Further, in one embodiment of the digital watermark detection processing method of the present invention, the standardization processing step is performed depending on whether a plurality of variances based on detection values of a plurality of digital watermark patterns are equal or different. It is characterized in that different t statistic calculation formulas are applied to execute the t statistic calculation processing based on the output of the variance comparison unit.

Further, a third aspect of the present invention is a program for executing a digital watermark pattern detection process on a computer system, wherein a plurality of digital watermarks are applied to an input image which is a target of the digital watermark pattern detection process. A digital watermark pattern detection value acquisition step of performing a detection process for each pattern to acquire a plurality of detection values for each digital watermark pattern, and a plurality of detections of each pattern acquired in the digital watermark pattern detection value acquisition step A standardization process step of performing a standardization process of the detection value difference of each digital watermark pattern based on the value, and obtaining a standardized value calculated as a result of the standardization process;
And a digital watermark embedding pattern determination step of acquiring embedded digital watermark information based on a comparison process between the standardized value acquired in the standardization processing step and a threshold value.

The program of the present invention is, for example, a medium provided in a computer-readable format to a general-purpose computer system capable of executing various program codes, for example, a storage medium such as a CD, FD, or MO. The program of the present invention can be stored in a medium that provides a computer program in a computer-readable format to a general-purpose computer system capable of executing various program codes. is there. The medium is not particularly limited in its form, such as a recording medium such as a CD, FD, MO or a transmission medium such as a network.

Such a program regulates the execution of various functions of the system based on the reading of the program under the control of the processor and exerts a cooperative action on the system. It is possible to obtain the same effect as that of the side surface.

The other objects, features and advantages of the present invention are as follows.
It will be clarified by a more detailed description based on embodiments of the present invention described below and the accompanying drawings.

[0128]

BEST MODE FOR CARRYING OUT THE INVENTION A digital watermark detection processing apparatus and a digital watermark detection processing method of the present invention will be described below.
A description will be given with reference to the drawings. The present invention integrates detection values of a plurality of digital watermarks, compares the detection values with a threshold value, and determines the presence or absence of a digital watermark and the embedded bit information based on the comparison result.

First, the outline of the digital watermark detection processing of the present invention will be described using the flow of FIG. The processing flow shown in FIG. 21 is an example of a case where bit information is obtained by detecting two digital watermark patterns. As will be described later, the configuration of the present invention can be a configuration in which three or more different digital watermark patterns are applied.

The processing flow of FIG. 21 will be described. In the digital watermark detection processing of the present invention, as shown in FIG.
First, the detection values of the two digital watermark patterns are independently obtained (S2101, S2102).

Subsequently, the detected values of the two digital watermark patterns are integrated, the integrated value is compared with the threshold, and it is determined whether it is equal to or more than the threshold or not (S210).
3). This integration processing is executed, for example, as processing for obtaining the difference value: d of the detection values of the respective digital watermark patterns and comparing the difference value: d with the threshold th. This determination processing will be further described later. Finally, the presence or absence of a digital watermark and the embedded bit information are determined based on the obtained determination result (S2104).

Conventionally, whether or not a digital watermark is embedded has been determined by comparing the absolute detection value of each digital watermark with a threshold value. For example, in FIG. 4, when the detection processing is performed on the original image, the detection values are distributed around the origin, and when the detection processing is performed on the image in which the digital watermark pattern is embedded, the detection values are cW · W. Can be expected to be distributed around.

This means that the distribution centers of the two detection values of the detection value of the original image and the detection value of the image in which the digital watermark pattern is embedded are separated by the difference cW · W. In embedding and detecting the digital watermark according to the present invention, the bit determination is performed using the difference between the detected values to determine the presence or absence of the digital watermark.

An example using two different digital watermark patterns 1 and 2 will be described below. A value obtained by subtracting the detection value x ₂ of the digital watermark 2 from the detection value x ₁ of the digital watermark ₁ is defined as a difference value: d indicating the difference between the detection values. That is, the detected value difference value: d is a value defined by the following equation.

[0135]

D = x ₁ −x ₂

For example, the electronic transparency described in the section of the prior art is used.
Figure showing bit information: 1 using patterns 1 and 2
In the example shown in FIG. 9A, the detection value of the digital watermark 1 is c₁W₁
・ W ₁, And the detection value of digital watermark 2 is
It is distributed around the origin. In this case, the detected value difference value:
d is d = c₁W₁・ W₁−0, the detection value difference value: d
Is c₁W₁・ W₁Will be distributed around.

Similarly, in the example shown in FIG. 9B in which bit information: 0 is expressed by using the digital watermark patterns 1 and 2,
The detection values of the digital watermark 1 are distributed around the origin, and the detection values of the digital watermark 2 are distributed around c ₂ W ₂ · W ₂ . In this case, the detected value difference value: d is d = 0-c ₂ W ₂ ·
W ₂ , and the detected value difference value: d is distributed around -c ₂ W ₂ · W ₂ .

[0138] Detection value x ₂ detection value difference value of the detected value of the electronic watermark pattern 1 x ₁ digital watermark pattern 2: as d = x ₁ -x _2, hereinafter, a description will be given of a specific detection processing example.
The distribution of the detection value difference value: d between the two digital watermark patterns 1 and 2 is distributed around d ₀ when bit information = 1 is embedded as shown in FIG. When embedded, it is distributed around -d _0, and when nothing is embedded, it is distributed around the origin. In order to make the explanation easier to understand,
was a _{d 0 = c 1 W 1 ·} W 1 = c 2 W 2 · W 2.

In such an aspect, bits 1, 0,
And the determination of the three modes in which the electronic watermark is not detected is performed based on the comparison between the detection value difference value: d = x ₁ -x ₂ and the threshold value.
That is, the determination process is performed based on the condition of the following equation.

[0140]

[Equation 21]

In FIG. 22, for easy understanding of the explanation, th ₁ = th th ₂ = -th is set for each threshold value.

[Bit Representation Method] When the presence or absence of a digital watermark and bit information are represented by the difference value d between two digital watermark detection values, a plurality of digital watermark detection values and threshold values are not used for comparison, but a plurality of digital watermark detection values are used. The difference value: d of the detected values of the digital watermark is compared with the threshold value. Therefore, even if the same bit information, for example, bit information = 1, various situations may occur in the detection mode of two digital watermark detection values. Some aspects thereof will be described below with reference to specific examples.

Various expression modes of bit information = 1 will be described. First, in the example shown in FIG. 23, the distribution of the detection value x ₁ of the digital watermark 1 shown in the upper row and the digital watermark 2 shown in the lower row
9 shows the case where the distribution of the detected values x ₂ of 1 is moved in the negative direction from the state of FIG. In this way, even if the distributions of the detection values of the digital watermarks 1 and 2 deviate in the negative direction,
By embedding a digital watermark so that the detected value difference value: d is held, the same bit information as in FIG.
1 can be detected.

It is also possible to further advance the movement in the negative direction to obtain the structure shown in FIG. In this case, the digital watermark pattern 1 is not embedded, but the digital watermark pattern 2 is embedded in the negative direction. Even if the digital watermark pattern is embedded in this way, it is possible to detect bit information = 1 as in FIG. 9A. FIG. 25 shows an example in which the moving strength in the negative direction is further increased.

In this way, the detection value difference value between the digital watermark pattern 1 detection value x ₁ and the digital watermark pattern 2 detection value x ₂ : d = x ₁ −x ₂ is th ₁ (= th) ≦ d When the digital watermark embedding process is executed by controlling the digital watermark embedding strength so that the detection values of the two digital watermark patterns are obtained so as to satisfy the condition, the bit information = 1 is determined in the digital watermark detecting process. It will be.

In the configuration of the present invention, the threshold value comparison between the individual detection values x ₁ and x ₂ of the digital watermark patterns ₁ and ₂ is not performed. Therefore, for example, although the original image data of the original image includes the same image data as the digital watermark pattern 2 and the digital watermark pattern 2 is not embedded, the correlation value similar to that of the digital watermark pattern 2 has a certain strength. Even if it is detected, the digital watermark embedding processing apparatus embeds the digital watermark pattern 2 in the negative direction so as to satisfy the condition of th ₁ (= th) ≦ d, or the other digital watermark pattern. By embedding 1 in the digital watermark detection processing device, the detection value difference value: d = x ₁ −x ₂
Can be acquired, and the judgment of bit information = 1 can be correctly executed.

On the other hand, the same processing can be performed for bit information = 0. FIG. 26 is an example of a representation method of bit information = 0, and the detection value difference value between the digital watermark pattern 1 detection value x ₁ and the digital watermark pattern 2 detection value x ₂ is: d =
x ₁ −x ₂ is an example in which a digital watermark is embedded by satisfying the condition of d ≦ th ₂ (= −th), and in this case, bit information = 0 is determined. Even in the digital watermark embedding process of bit information = 0, various modes in which deviations in various directions are generated are possible as in the case of bit information = 1 described above.

In short, the determination of the presence or absence of a digital watermark and the determination of bit information are not determined by the absolute distance from the origin of the detection value of each digital watermark, but to the detection values of a plurality of digital watermarks. It will be determined by the relative positional relationship.

[Multi-bit Digital Watermark Pattern Representation] In the above description, two digital watermark patterns are used to represent 1-bit information. However, the present method is not limited to this, and can be applied even when three or more digital watermark patterns are used to represent 1-bit information. That is, this method evaluates the detected value of each digital watermark pattern according to each relation.

In FIG. 27, 3 is used to represent 1-bit information.
Shows an example using one digital watermark pattern 1, 2, 3
You As shown in FIGS. 27A to 27C, three digital watermarks
Detection value x of each of patterns 1, 2 and 3₁, X₂, X₃Nou
If there are two, and one is apart in the positive direction,
When it can be judged, it is embedded in the image with a digital watermark.
Bit information present = 1. Also, as shown in FIG.
To detect each of the three digital watermark patterns 1, 2, 3
Value x₁, X₂, X₃Two of them are available and only one is negative
When it can be judged that they are separated in the direction of
It is assumed that the embedded bit information = 0. FIG. 27
FIG. 27D is a shape obtained by just inverting FIG. In the figure
Although not shown, similarly, an inverted form of FIGS. 27 (b) and 27 (c).
, The bit information embedded in the digital watermark is
Report = 0. In all other cases, that is, three electronic transparency
Detection value x of each of the pattern 1, 2, 3₁, X₂, X₃
If you don't have all of the
It is determined that no digital watermark is embedded. That is, in advance
With respect to the determined threshold th, the digital watermark pattern 1
Detection value x₁, Detection value x of digital watermark pattern 2₂, Electronic transparent
Detected value x of the pattern 3 ₃And bit judgment
Is determined according to the following conditional expression.

[0151]

[Equation 22]

Under such a condition, the embedding strength is controlled to embed the digital watermark, and the detection processing device performs the detection under the above-mentioned determination condition, whereby a highly reliable electronic device is obtained. Enables watermark embedding and detection. The digital watermark embedding processing apparatus embeds one digital watermark selected from a plurality of digital watermark patterns or adjusts the embedding strength of each digital watermark pattern so as to satisfy the above conditional expression. To embed multiple digital watermarks.

Further, an example of expressing information of 2 bits or more by using three or more digital watermark patterns will be described with reference to FIG.

In FIG. 28, 3 is used to represent 2-bit information.
An example using one digital watermark pattern 1, 2, 3
It As shown in FIG. 28A, the digital watermark pattern 1 is detected.
Outgoing value x ₁, The detected values of the digital watermark patterns 2 and 3
x₂, X₃If it can be determined that is in the negative direction,
Bit information embedded in image by blank = 00
It However, the numerical values are displayed in binary. In addition, FIG.
As shown in (b), the detected value x of the digital watermark pattern 1₁When
Detection value x of digital watermark pattern 3₃Are available, electronic
Detection value x of watermark pattern 2₂Only the negative threshold t
If the distance is more than h, the bit information is 01.

As shown in FIG. 28C, the detection value x ₁ of the digital watermark pattern ₁ and the detection value x ₃ of the digital watermark pattern ₃ are aligned, and the detection value x ₂ of the digital watermark pattern 2 is obtained.
If only those are separated by a threshold value th or more in the positive direction, bit information = 10. Further, as shown in FIG. 28D, with respect to the detection value x ₁ of the digital watermark pattern ₁ , the detection values x ₂ and x _{3 of} the digital watermark patterns ₂ and ₃ are the threshold value th in the positive direction.
When the distance is larger than the above, bit information = 11. That is, the detection value of the digital watermark pattern 1 is defined as a higher bit and the detection value of the digital watermark pattern 3 is expressed as a lower bit with reference to the detection value of the digital watermark pattern 1.

In the example shown in FIG. 28, a predetermined threshold value t
For h, the detection value x _{1 of} the digital watermark pattern ₁ , the detection value x _{2 of} the digital watermark pattern ₂ , the digital watermark pattern 3
When the detected value x is set to x ₃ , the bit determination is made according to the following conditional expression.

[0157]

[Equation 23]

Under such a condition, the embedding strength control is executed to embed the digital watermark, and the detection processing device executes the detection under the above-mentioned judgment condition, whereby a high reliability is obtained. Digital watermark embedding and detection of bit information is possible. The digital watermark embedding processing apparatus embeds one digital watermark selected from a plurality of digital watermark patterns or adjusts the embedding strength of each digital watermark pattern so as to satisfy the above conditional expression. To embed multiple digital watermarks.

[Embedding Process Based on Detection Method Using Detection Value Difference] Next, it is premised that the bit value determination based on the difference value d between the detection values of the plurality of digital watermark patterns as described above is executed. As an example, a digital watermark embedding processing apparatus that executes a digital watermark embedding process will be described with reference to FIG.

The digital watermark pattern generation unit 2004 generates a digital watermark pattern from the embedded information 2002 embedded in the image and the digital watermark pattern generation key (key) in the digital watermark pattern generation key (key) storage unit 2003.

The embedded information 2002 is information embedded as a digital watermark, and is arbitrary information such as copy control information, copyright information, and edit information. Specifically, the digital watermark pattern generation key (key) corresponds to the method of dividing the image into small areas described in FIG. 6 when embedding the digital watermark pattern in the image, and the bits of the embedded information described in FIG. 7. Image division information, bit array information, etc., such as how to allocate small areas to be embedded, and processing information necessary to generate the embedded information 2002 as a digital watermark pattern.

In the digital watermark detecting section 2005, the original image 2
The detection processing is performed on 001 to check how many digital watermark components are potentially included in the original image, and the result and the original image are passed to the electronic watermark pattern embedding unit 2006.

In the digital watermark pattern embedding unit 2006, the digital watermark pattern generated by the digital watermark pattern generating unit 2004 is added to the original image while increasing or decreasing the embedding strength based on the result obtained by the digital watermark detecting unit 2005. Embed in 2001. The image in which the digital watermark pattern is embedded is output as a digital watermark embedded image 2007.

A concrete mode of the digital watermark embedding process executed in the digital watermark embedding process of FIG. 29 is shown in FIG.
It demonstrates using 0-FIG. 30 to 33, the detection value x ₁ ′ of the digital watermark ₁ and the detection value x ₂ ′ of the digital watermark 2 with respect to the original image (image before embedding the digital watermark)
Is represented by a solid arrow. The detected value x ₁ ′ and the detected value x ₂ ′ indicate the inclusion state of the correlation component with the digital watermark 1 and the digital watermark 2 originally included in the image component of the original image. The dotted arrow
For example, it represents the embedding strength of a digital watermark pattern executed in the digital watermark embedding processing apparatus shown in FIG. Detection value x ₁ 'of digital watermark ₁ and detection value x ₂ ' of digital watermark 2 with respect to the original image (image before digital watermark embedding)
Is data acquired by the digital watermark detection unit 2005 of the digital watermark embedding processing apparatus shown in FIG.

It is assumed that the detection value of the digital watermark 1 from the image after embedding the digital watermark pattern is x ₁ and the detection value of the digital watermark 2 is x ₂ . 30 to 33 are all bit information =
1 and the detection intensities d = x ₁ −x ₂ are equal. 30 to 32, the potential digital watermark component of the original image, that is, the digital watermark detection strength detected from the original image itself is the same, but the method of embedding the digital watermark pattern is different.

In the example shown in FIG. 30, the digital watermark pattern 1 is used.
This is an example of a digital watermark embedding process in which the difference value of the detection values of the digital watermark 1 and the digital watermark 2: d = x ₁ −x ₂ is separated by a threshold value th or more by embedding only the above. That is, the detection value x of the digital watermark 1 with respect to the original image (the image before the digital watermark embedding) acquired by the digital watermark detection unit 2005 of the digital watermark embedding processing device shown in FIG.
_{Based on 1} ′ and the detection value x ₂ ′ of the digital watermark 2, the digital watermark pattern embedding unit 2006 sets the embedding strength of the digital watermark 1 so that x ₁ −x ₂ > th. The embedding process of the digital watermark pattern 1 is executed with the strength indicated by the dotted arrow. As a result of this embedding processing, the digital watermark detection processing apparatus calculates d = x ₁ −x ₂ > th for the difference value d between the two digital watermark pattern detection values.
Is determined, and based on this determination, it is concluded that the digital watermark indicates bit = 1.

In the example shown in FIG. 31, the digital watermark pattern 2 is used.
This is an example of the digital watermark embedding process in which the difference value d = x ₁ -x ₂ between the detection values of the digital watermark 1 and the digital watermark 2 is separated by a threshold value th or more by embedding only the negative strength in the negative direction. That is, the detection value x ₁ ′ of digital watermark ₁ and the detection value x _{2 of} digital watermark 2 with respect to the original image (image before digital watermark embedding) acquired by the digital watermark detection unit 2005 of the digital watermark embedding processing device shown in FIG. 29. ', The digital watermark pattern embedding unit 2006 determines the embedding strength of the digital watermark 2 as x ₁ −x ₂ > th.
Then, the embedding process of the digital watermark pattern 2 is executed with the negative intensity indicated by the dotted arrow in FIG. As a result of this embedding process, the digital watermark detection processing device
Regarding the difference value d between the two digital watermark pattern detection values,
The judgment of d = x ₁ −x ₂ > th is made, and based on this judgment, it is concluded that the digital watermark indicates bit = 1.

In the example shown in FIG. 32, the digital watermark pattern 1 is used.
Is embedded with the intensity in the positive direction and the digital watermark pattern 2 is embedded with the intensity in the negative direction, and the difference value between the detected values of the digital watermark 1 and the digital watermark 2: d = x ₁ -x _{2 is set} to the threshold value t.
This is an example of a digital watermark embedding process in which the process of separating from h is performed. That is, the digital watermark 1 for the original image (image before digital watermark embedding) acquired by the digital watermark detection unit 2005 of the digital watermark embedding processing apparatus shown in FIG.
Based on the detection value x ₁ ′ of the above and the detection value x ₂ ′ of the digital watermark 2 in the digital watermark pattern embedding unit 2006,
The embedding strength of the digital watermark 1 and the embedding strength of the digital watermark 2 are set so that x ₁ −x ₂ > th, and the embedding process of the digital watermark pattern 1 is executed at the strength indicated by the dotted arrow in FIG. 32. The embedding process of the digital watermark pattern 2 is executed with the negative intensity indicated by the dotted arrow in FIG. As a result of this embedding process, the digital watermark detection processing device
Regarding the difference value d between the two digital watermark pattern detection values,
The judgment of d = x ₁ −x ₂ > th is made, and based on this judgment, it is concluded that the digital watermark indicates bit = 1.

In the example shown in FIG. 33, similar to FIG. 32, the digital watermark pattern 1 is embedded with the intensity in the positive direction and the digital watermark pattern 2 is embedded with the intensity in the negative direction. Difference value of detected values: d = x
_This is an example of a digital watermark embedding process in which _1- x ₂ is separated by a threshold value th or more. However, the original image acquired by the digital watermark detection unit 2005 of the digital watermark embedding processing apparatus shown in FIG. The detection value x ₁ ′ of the digital watermark ₁ and the detection value x ₂ ′ of the digital watermark 2 for the image are different from those in the case of FIG. 32, and x ₁ −x ₂ > th
Therefore, this is an example of a case where the embedding strength of each digital watermark needs to be greatly adjusted.

In the digital watermark pattern embedding unit 2006, the embedding strength of the digital watermark 1 and the digital watermark 2
33 is set so that x ₁ −x ₂ > th, the embedding process of the digital watermark pattern 1 is executed with the intensity shown by the dotted arrow in FIG. 33, and the negative intensity shown by the dotted arrow in FIG. The process of embedding the digital watermark pattern 2 is executed. As a result of this embedding processing, the digital watermark detection processing device makes a determination of d = x ₁ −x ₂ > th for the difference value d between the two digital watermark pattern detection values, and indicates bit = 1 based on this determination. Get the conclusion that it is a digital watermark.

As described above, the digital watermark embedding processing apparatus obtains the detection value of each digital watermark pattern in the original image, and sets at least one digital watermark embedding strength so that the difference value d after embedding becomes equal to or greater than the threshold th. Adjust and execute the embedding process. However, if the embedding strength of either one of the digital watermark patterns becomes extremely large, the image quality is likely to deteriorate, so it is desirable to appropriately allocate the two digital watermark patterns. The allocation is
Various methods are conceivable, such as halving the embedding strength, making the detection values equal, or making it proportional to the potential digital watermark component of the original image.

FIG. 34 shows a digital watermark embedding processing flow in the digital watermark embedding processing apparatus. Each step of the processing flow of FIG. 34 will be described.

In step S2501, the original image which is the target of the digital watermark embedding process is input. Step S25
In 02, with respect to the input original image, the degree of correlation between the original image before embedding the electronic watermark and each electronic watermark pattern applied to the embedding process is checked. That is, the digital watermark pattern 1 and the digital watermark pattern 2 are detected in the original image. Although this flow describes an example in which two digital watermark patterns are applied, when three or more digital watermark patterns are applied, detection processing of each of those digital watermark patterns is executed.

The digital watermark pattern 1 and the digital watermark pattern 2 are detected from the original image, and the obtained detection values are set as x ₁ 'and x ₂ '.

[0175] In step S2503, determines the difference _{d '= x 1' -x 2} ' of the detection value. In step S2504, the total embedding strength is determined by comparing the detected value difference d ′ with the threshold th. In step S2505, the total embedding strength of the digital watermark pattern is assigned to the digital watermark pattern 1 and the digital watermark pattern 2 based on the detection value x ₁ ′ of the digital watermark ₁ and the detection value x ₂ ′ of the digital watermark 2 from the original image. . As described above, the allocation is a process of embedding only one digital watermark, halving the embedding strength, making the detected values equal, or making it proportional to the potential digital watermark component of the original image. And various methods are feasible. Step S2506
Then, the digital watermark pattern 1 and the digital watermark pattern 2 are embedded in the original image according to the determined embedding strength.
In step S2507, the image in which the digital watermark pattern is embedded is output and the process ends.

[Embedding Strength Allocation Method] In step S2505 of the above-described processing flow, the processing of allocating the total embedding strength of the digital watermark to the digital watermark pattern 1 and the digital watermark pattern 2 is executed. It is possible to analyze the image that is the target of the digital watermark embedding process and execute the process based on the analysis result.

A processing example of setting the embedding strength of each digital watermark pattern by image analysis will be described with reference to FIG. In this example, the embedding strength of each digital watermark pattern is determined according to the visual characteristics of the image.

For example, it is assumed that the embedding areas of the digital watermark patterns W ₁ and W ₂ are set as shown in FIG. The area where the digital watermark pattern W ₁ is embedded is an empty portion in the image, and the pixel value is flat (area that does not contain high frequency components). On the contrary,
The area of the digital watermark pattern W ₂ is a forest portion in the image, and the pixel value is random (area containing a large amount of high frequency components).

Therefore, in the area of the digital watermark pattern W ₁ , it is difficult to embed the digital watermark pattern so that it is difficult for a person to recognize the digital watermark pattern, and the embedding strength of the digital watermark pattern must be suppressed. on the other hand,
In the area of the digital watermark pattern W ₂ , the original image contains many high frequency components, and it is easy to embed the digital watermark pattern so that it is difficult for humans to recognize the digital watermark pattern. Can be relatively strong.

In the method of detecting a digital watermark by the difference between the detected values of two digital watermarks, the digital watermark pattern 1 and the digital watermark pattern 2 satisfy the condition that the difference d between the detected values is larger than the threshold th. Good luck
The strength of each pattern can be set freely, and either of them can be embedded. Therefore, the embedding strength of the digital watermark pattern 1 that is difficult to embed is reduced, and the digital watermark pattern 2 that is easy to embed is used to reduce the embedding strength. It is possible to suppress deterioration of image quality without changing the detection intensity.

That is, when embedding the digital watermark pattern 1 and the digital watermark pattern 2 in the image shown in FIG. 35, the embedding strength of the digital watermark pattern 1 is weakened or the embedding process is not performed, while the digital watermark pattern 2 is embedded. By increasing the embedding strength of and the difference is set to be larger than the threshold value, it becomes possible to embed a digital watermark pattern that is difficult for human eyes to see. For example, by embedding only the digital watermark pattern 2 with negative strength and positive strength, for example, if d = x ₁ −x ₂ ≧ th for the difference value d of each digital watermark pattern, bit information = 1, If d = x ₂ −x ₁ ≧ th, it is possible to embed and detect each bit information so that bit information = 0.

FIG. 36 shows a processing flow for executing the digital watermark embedding processing by setting the embedding strength of each digital watermark pattern by image analysis. Each step of the processing flow of FIG. 36 will be described.

In step S2701, an original image to be digital watermark embedding processing target is input. Step S27
In 02, with respect to the input original image, the degree of correlation between the original image before embedding the electronic watermark and each electronic watermark pattern applied to the embedding process is checked. That is, the digital watermark pattern 1 and the digital watermark pattern 2 are detected in the original image. Although this flow describes an example in which two digital watermark patterns are applied, when three or more digital watermark patterns are applied, detection processing of each of those digital watermark patterns is executed.

The detection of the digital watermark pattern 1 and the digital watermark pattern 2 is performed on the original image, and the respective detected values obtained are defined as x ₁ 'and x ₂ '.

[0185] In step S2703, determines the difference _{d '= x 1' -x 2} ' of the detection value. In step S2704, the total embedding strength is determined by comparing the detection value difference d ′ with the threshold th. In step S2705, the image of the area in which the digital watermark pattern 1 and the digital watermark pattern 2 are embedded is checked to check the embedding allowable strength of the digital watermark pattern, that is, the strength that further deterioration cannot be denied. In general, it can be used that the change of the pixel value of the edge part is difficult to recognize, but the change of the pixel value of the flat part is easy to recognize. In addition, various image evaluation methods used when embedding a digital watermark can be used. For example, the brightness is high or low, and if it is a moving image, it is the amount of movement.

In step S2706, the detected value x ₁ ′ of the digital watermark ₁ and the detected value x ₂ ′ of the digital watermark 2 from the original image.
Then, the total embedding strength of the digital watermark pattern is assigned to the digital watermark pattern 1 and the digital watermark pattern 2 based on the embedding allowable strength obtained in step S2705. As described above, the allocation is a process of embedding only one digital watermark, halving the embedding strength,
Various methods are possible, such as making the detection values equal, or making them proportional to the potential watermark component of the original image. In step S2707, the digital watermark pattern 1 and the digital watermark pattern 2 are embedded in the original image according to the determined embedding strength. Step S270
In step 8, the image in which the digital watermark pattern is embedded is output and the process ends.

[Detection Process Performing Detection Method Using Detection Value Difference] Next, a bit value determination or a digital watermark presence / absence determination is performed based on the difference value d of the detection values of each of the plurality of digital watermark patterns. A digital watermark detection processing device that does this will be described with reference to FIG.

The digital watermark pattern generation unit 3003 generates a digital watermark pattern from the digital watermark pattern generation key (key) in the digital watermark pattern generation key (key) storage unit 3002. The digital watermark pattern generation unit 3003 generates a plurality of digital watermark patterns to detect.

The digital watermark pattern generation key (key) is
Specifically, it is image division information when embedding a digital watermark pattern in an image, bit array information, and the like, and is information necessary for detecting a digital watermark pattern.

The digital watermark detection value acquisition unit 3011 of the detection unit 3004 detects the digital watermark of the input image 3001 using each digital watermark pattern generated by the digital watermark pattern generation unit 3003, and detects each digital watermark pattern. The detection value x _m for

With respect to the detection value acquired by the detection value acquisition unit 3011, the detection value difference acquisition unit 3012 calculates the difference between the detection values of the respective digital watermarks, for example, d = x _m-1 −x _m .
The difference data d is detected by the detection value difference determination unit 3013.
It is compared with a predetermined threshold value, and bit information is acquired according to a predetermined determination condition. The determination condition is, for example, each of the above [Formula 21], [Formula 22], and [Formula 23]. The information detected by the detection unit 3004 is output as detection information 3005.

FIG. 38 shows a digital watermark detection processing flow in the digital watermark detection processing apparatus of the present invention. Each step of the processing flow of FIG. 38 will be described.

In step S3101, an image to be subjected to digital watermark detection processing is input. In step S3102, digital watermark pattern 1 and digital watermark pattern 2 are detected in the input image. Although this flow describes an example in which two digital watermark patterns are applied, when three or more digital watermark patterns are applied, detection processing of each of those digital watermark patterns is executed.

Detection of the digital watermark pattern 1 and the digital watermark pattern 2 is performed on the input image, and the obtained detection values are set as x ₁ and x ₂ . In step S3103, it determines the difference d = x ₁ -x ₂ of this detection value. In step S3104, the difference d between the detected values and the threshold th are compared to determine bit information. The determination conditions are, for example, the above-mentioned [Formula 21] and [Formula 2]
2] and [Equation 23]. Finally, step S
At 3105, the detected information is output as detection information.

[Standardization of Two-Sample Problem (Difference of Sample Mean)]
Next, in the digital watermark embedding process and the detection process for acquiring bit information using the plurality of digital watermark detection value difference values: d described above, standardization processing is applied to embed digital watermarks in various images. A configuration for improving the detection accuracy in the execution configuration of the detection process will be described.

In the method described above, the bit determination processing is performed by applying the difference value: d of a plurality of digital watermark patterns. The digital watermark pattern detection processing is executed in a plurality of image areas in one image, as shown in FIG. 5, FIG. 6, FIG. 7, or FIG.
Furthermore, in a moving image, as described with reference to FIG. 17, it may be continuously executed in a plurality of continuous image frames. That is, many detection values are acquired for each of the plurality of digital watermark patterns.
A more stable bit determination process can be performed by comprehensively determining a large number of these detection values. One of the methods is standardization processing that considers dispersion.

For example, assuming that a large number of detection values of each digital watermark pattern are acquired using two digital watermark patterns 1 and 2, the variance of each detection value will be considered. The average value of the detection values of the digital watermark pattern 1 is μ ₁ , the variance is σ ₁ ² , the average value of the detection values of the digital watermark pattern 2 is μ ₂ , and the variance is σ ₂ ² .

Random variable X according to mean value μ ₁ and variance σ ₁ ²
Consider ₁ and the random variable X ₂ according to the mean value μ ₂ and the variance σ ₂ ² . The average value [X ₁ ] of the random variable X ₁ is defined by the following equation.

[0199]

[Equation 24]

The average value [X ₂ ] of the random variable X ₂ is defined by the following equation.

[0201]

[Equation 25]

The mean value [X ₁ ] of each of these random variables,
If [X ₂ ] is large enough for sample sizes n ₁ and n ₂ ,
Normal distributions N (μ ₁ , σ ₁ ² / n ₁ ), (μ ₂ , σ ₂ ²⁾
/ N ₂ ) is shown by the central limit theorem. However, if the distributions of the random variables X ₁ and X ₂ are normal distributions, the distributions of the average values [X ₁ ] and [X ₂ ] are exactly normal distributions. Here, n ₁ and n ₂ are natural numbers.

The difference between these average values [X ₁ ] and [X ₂ ] is obtained by the following equation.

[0204]

[Equation 26]

Consider the difference between the mean values [X ₁ ] and [X ₂ ] of each random variable defined by the above equation. If the population variances σ ₁ ² and σ ₂ ^{2 of the two} populations are known, the difference [X
The distribution of ₁ ] − [X ₂ ] becomes a normal distribution N (μ ₁ −μ ₂ , (σ ₁ ² / n ₁ ) + (σ ₂ ² / n ₂ )) if n ₁ and n ₂ are sufficiently large. The standardized value Z is shown by the following formula.

[0206]

[Equation 27]

The standardized value Z defined by the above equation is
The standard normal distribution is N (0,1). However, if the distributions of the random variables X ₁ and X ₂ are normal distributions, the distribution of the standardized value Z is exactly the standard normal distribution.

A process of detecting embedded information based on the standardized value Z will be described with reference to FIG. In FIG. 39, the horizontal axis represents the standardized value Z. On the basis of the standardized data Z of the plurality of detection value data of the digital watermark embedding pattern, the bit information 0 or 1 of the digital watermark or the judgment when the digital watermark is not detected is based on the following conditional expression. .

[0209]

[Equation 28]

However, when the standardized value Z is actually applied to the detection of the digital watermark, the population variances σ ₁ ² and σ ₂ ² corresponding to the digital watermark patterns 1 and ² are unknown.
The standardized value Z of [Equation 27] cannot be applied unconditionally. Corresponding to the digital watermark patterns 1 and 2 when the mother variances σ ₁ ² and σ ₂ ² corresponding to the digital watermark patterns 1 and ² are equal, that is, σ ₁ ² = σ ₂ ² = σ ² When the population variances σ ₁ ² and σ ₂ ² are not equal, that is, σ ₁ ² ≠ σ ₂ ²
If it is, consider it separately. However, when the population variances σ ₁ ² and σ ₂ ² are known by some means, [Equation 27] can be used as it is.

The population variances σ ₁ ² and σ ₂ ² are unknown, but σ ₁ ² =
First, the processing when σ ₂ ² = σ ² is described. Random variable X ₁ according to mean value μ ₁ , variance σ ₁ ² and mean value μ ₂ ,
Difference [X ₁ ]-[X of mean values of random variables X ₂ according to variance σ ₂ ²
₂ ] of the sample variance s ² is defined by the following equation.

[0212]

[Equation 29]

The sample variance s defined by the above equation
^In the case of substituting the population variance σ ^{2 in 2} , [Equation 27] can be rewritten as the following equation.

[0214]

[Equation 30]

This is a 2-sample t statistic, and has n ₁ degrees of freedom.
The distribution is + n ₂ -2.

Next, the population variances σ ₁ ² and σ ₂ ² are unknown, and σ
_The case where ₁ ² ≠ σ ₂ ² is shown. In this case, the population variances σ ₁ ² and σ
₂ ² Welch's approximation is known, although the statistics that are not related to ² are uncertain. Where the sample variances s ₁ ² and s
₂ ² is defined by the following formula.

[0217]

[Equation 31]

The sample variance s ₁ ² defined by the above equation,
When the population variances σ ₁ ² and σ ₂ ² are substituted by s ₂ ² , [Welch's approximation method] can be rewritten as [Equation 27] into the following equation.

[0219]

[Equation 32]

It is known that the above equation is obtained and the degree of freedom φ approximately follows the t distribution defined by the following equation.

[0221]

[Expression 33]

In the digital watermark detection process of the present embodiment, the detection values x ₁ and x ₂ of the digital watermark patterns 1 and 2 are used as a distribution serving as a criterion for determining whether or not a plurality of digital watermark patterns to be detected are embedded. Instead of using the distribution,
We will use the distribution of t statistics. [Numerical equation 30] described above
Alternatively, the t statistic calculated by [Equation 32] is compared with a threshold value, and the determination of bit information 0 and 1 and the presence / absence of a digital watermark are performed according to the determination condition of [Equation 28].

Here, in order to obtain one t statistic according to each of the above equations using the detection value obtained from one frame of the image, it is clear from the definitions of [Equation 24] and [Equation 25] that each detection It is necessary to have n ₁ and n ₂ values x ₁ and x ₂ , respectively. As one method of acquiring these numbers of digital watermark pattern detection values, for example, one frame of an image is divided into small regions as shown in FIG. 14 described above, and a large number of detection values are obtained from one frame. There is. In FIG.
Electronic watermark pattern 1: Embedding j number of W ₁ in the leftmost column, right next to the electronic watermark pattern in column 2: the W ₂ j number, electronic watermark pattern in the rightmost column i: the W _i to j number embed Example Equivalent to. Note that although the grid is divided in FIG. 14, a division other than the grid may be used. FIG. 14 shows an example in which a large number of i types of digital watermark patterns are embedded in one frame. When embedding only two types of digital watermark patterns, only the digital watermark pattern 1 and the digital watermark pattern 2 may be dispersed and embedded in each small area.
When the digital watermark pattern 1 and the digital watermark pattern 2 are detected, the average value [x ₁ ] of the detection values x ₁ and x _{2 in} small area units,
[X ₂ ] and variances s ₁ ² and s ₂ ² are calculated. Since the population mean μ ₁ and μ ₂ of the distribution of the detected values x ₁ and x ₂ with respect to the original image are μ ₁ = 0 and μ ₂ = 0 from the definition of the digital watermark, the t statistic has a variance of σ ₁ ² = When σ ₂ ² , the following formula is obtained.

[0224]

[Equation 34]

When the variance is σ ₁ ² ≠ σ ₂ ² , the t statistic is given by the following equation.

[0226]

[Equation 35]

The t distribution shown in [Equation 34] has a degree of freedom n ₁
Since it is uniquely determined by + n ₂ −2, the number of small areas n ₁ ,
By determining n ₂ , the obtained t distribution is uniquely determined.
Further, since the t distribution shown in [Equation 35] has a degree of freedom approximately determined by φ in [Equation 33], the distribution depends on the sample variance. However, the t distribution is uniquely determined when the degree of freedom is obtained.

From the above, the distribution used for judging the detection of the digital watermark becomes stable for any original image. However, as is clear from the definition, the variances s _1, ² and s
₂ ^{2 _{s 1 2 = s 2 2 =}} 0 become possible is not permitted. Therefore, when the variance s ₁ ² = s ₂ ² = 0, exceptional processing is required.

[Standardization of Two-Sample Problem (Ratio of Sample Variance)]
As described above, when the population variances σ ₁ ² and σ ₂ ² corresponding to the digital watermark patterns 1 and ² are unknown, σ ₁ ² = σ ₂ ² when they are equal, and σ ₁ ² ≠ when they are not equal. With σ ₂ ²
The method of obtaining the t statistic is different. In actual calculation, either method will be selected. As an example of the selection method, the degree of freedom (n ₁ −
It can be used to follow the F distribution of 1, n ₂ −1). The F distribution is a distribution that is uniquely determined by the degree of freedom.

[0230]

[Equation 36]

Specifically, the null hypothesis σ₁ ²= Σ₂ ², Conflict
Theory σ₁ ²≠ σ₂ ²Set the significance level under
Perform a test. If the null hypothesis is rejected, then σ₁ ²
≠ σ ₂ ²Since there is a high possibility that
Then, the t statistic is obtained. Null hypothesis rejected
If not, σ₁ ²= Σ₂ ²Is likely to be
The t statistic is calculated by [Equation 30]. This procedure
Is generally called the F test.

Information detection from the digital watermark pattern is executed by executing a process of determining the t statistics and the degree of freedom of the t distribution depending on whether the population variances of the two digital watermark detection values are equal. FIG. 40 shows the configuration of a digital watermark detection processing device that does this.

The digital watermark pattern generation unit 4003 generates a digital watermark pattern from the digital watermark pattern generation key (key) of the digital watermark pattern generation key (key) storage unit 4002. The digital watermark pattern generation unit 4003 generates a plurality of digital watermark patterns to detect.

A digital watermark pattern generation key (key) is
Specifically, it is image division information when embedding a digital watermark pattern in an image, bit array information, and the like, and is information necessary for detecting a digital watermark pattern.

The digital watermark detection value acquisition unit 4011 of the detection unit 4004 detects the digital watermark of the input image 4001 by using each digital watermark pattern generated by the digital watermark pattern generation unit 4003, and detects each digital watermark pattern. The detection value x _m for

The detection value acquired by the detection value acquisition unit 4011 is input to the detection value standardization processing unit 4012. The detection value standardization processing unit 4012 performs standardization processing of the detection value difference of each digital watermark pattern based on the plurality of detection values of each pattern acquired by the digital watermark pattern detection value acquisition unit 4011, and executes the standardization processing of the standardization processing. Obtain the resulting standardized value.

The variance comparison unit 4021 of the detection value standardization processing unit 4012 determines the degree of variance approximation of the detection values of the digital watermark pattern 1 and the digital watermark pattern 2. This determination process is performed based on a plurality of detection value results of the digital watermark pattern 1 and the digital watermark pattern 2 acquired by the detection value acquisition unit 4011. The variance comparison unit 4021 obtains n or more predetermined detection value data for each of the digital watermark pattern 1 and the digital watermark pattern 2 and generates a mother of each detection value of the digital watermark pattern 1 and the digital watermark pattern 2. The similarity of the variances is determined, and the mother variances of the detected values of the digital watermark pattern 1 and the digital watermark pattern 2 are equal, that is, σ
Whether ₁ ² = σ ₂ ² may be judged or not, that is, σ ₁ ² ≠ σ ₂ ²
It is determined whether or not the result is output to the t statistic calculation unit 4022. The variance comparison unit 4021 has analog variance determination reference data of the mother variance of the detection value of the digital watermark pattern, and determines whether the mother variances are equal according to this criterion.

In the t statistic calculation unit 4022, the mother variances of the detected values of the digital watermark pattern 1 and the digital watermark pattern 2 from the variance comparison unit 4021 are equal, that is, σ ₁ ² = σ ₂ ²
When the determination result of is input, the t statistic is obtained by applying the above [Equation 30], and the degree of freedom of the t distribution is n ₁ + n ₂
Set to -2 to define t distribution. On the other hand, the variance comparison unit 40
21 to digital watermark pattern 1 and digital watermark pattern 2
When the population variances of the detected values of are not equal, that is, when the determination result of σ ₁ ² ≠ σ ₂ ² is input, the t statistic is obtained by applying the above [Formula 32], and the degree of freedom of t distribution The above [Equation 3
3], the t distribution is approximately determined.

The t statistic calculating unit 4022 outputs the calculated t statistic and t distribution to the detection value difference determining unit 4013, and the detection value difference determining unit 4013 determines the threshold value and the t statistic determined by the input t distribution. The comparison is performed, and the presence or absence of a digital watermark and the bit information when the digital watermark is detected are determined based on the comparison result. Information detected by the detection unit 4004 by these processes is output as detection information 4005.

FIG. 41 shows a digital watermark detection processing flow in the digital watermark detection processing apparatus of the present invention. This Figure 4
The process indicated by 1 is a process executed by the detection unit 4004 of the digital watermark detection processing device shown in FIG. 40 described above. Each step of the processing flow of FIG. 41 will be described.

In step S4101, an image to be subjected to digital watermark detection processing is input, and with respect to the input image,
The digital watermark pattern 1 and the digital watermark pattern 2 are detected. Although this flow describes an example in which two digital watermark patterns are applied, when three or more digital watermark patterns are applied, detection processing of each of those digital watermark patterns is executed.

In step S4102, variances are independently obtained from the detection values of digital watermark 1 and digital watermark 2. In step S4103, it is tested whether it is possible to determine that the population variances of the detected values of the digital watermark 1 and the digital watermark 2 are equal. If both can be determined to be equal, the process moves to step S4104. In step S4104, the t statistic is calculated by [Equation 30]. In step S4105, the degree of freedom of the t distribution is obtained and the t distribution is determined. Step S410
At 8, the threshold value determined by the t distribution is compared with the t statistic.
Then, in step S4109, the presence or absence of a digital watermark and the bit information when the digital watermark is detected are determined based on the comparison result.

If it is not determined in step S4103 that the mother variances of the detected values of digital watermark 1 and digital watermark 2 are equal, the process advances to step S4106. Step S
At 4106, the t statistic is calculated by [Equation 32]. In step S4107, the degree of freedom of the t distribution is calculated as [Equation 33], and the t distribution is approximately determined. In step S4108, the t statistic is compared with the threshold determined by the t distribution. Then, in step S4109, the presence or absence of a digital watermark and the bit information when the digital watermark is detected are determined based on the comparison result.

[Dispersion Mode and Examples of Standardization Processing] The following will show examples of how the magnitude of dispersion of the digital watermark pattern 1 and the digital watermark pattern 2 changes due to standardization. 42 shows the case where both the digital watermark pattern 1 and the digital watermark pattern 2 show a small dispersion, and FIG. 43 shows the case where the digital watermark pattern 1 and the digital watermark pattern 2 are shown.
Both show large dispersion. Further, FIGS. 44 and 45 show the case where one has a small dispersion and the other has a large dispersion. Here, 2
It is assumed that the embedded bit information by the combination of two digital watermark patterns is bit information = 1. Also fals
The threshold th so that the probability p _{FP of} e positive becomes constant
_Let's decide _Z.

As shown in FIG. 42, when both digital watermark pattern 1 and digital watermark pattern 2 have a small variance,
The standardized value Z will be distributed around a point that is far away from the origin 0. That is, the center of the distribution and the threshold th
The difference d _Z of _Z is increased, the probability becomes a false negative p _FN
It can be seen that the result can be suppressed to a low level and the detection result can be expected to be reliable.

On the other hand, as shown in FIG. 43, when both the digital watermark pattern 1 and the digital watermark pattern 2 have large dispersion, the standardized value Z is distributed around the origin 0. That is, the center of the distribution and the threshold th _Z
It is possible that the difference d _Z between the two becomes smaller and the center of the distribution falls below the threshold th _Z in some cases. In this case, false
probability p _FN to be a negative is high, the reliability it can be seen that that can not be expected on the detection result. At this time, it is preferable to increase the number of samples in order to reduce the dispersion. For example, in the case of a moving image, a detection value obtained from another frame is used.

Further, as in the examples shown in FIGS. 44 and 45,
When the digital watermark pattern 1 and the digital watermark pattern 2 have different sizes of dispersion, the standardized value Z has an intermediate behavior between those in FIGS. 42 and 43. That is, the difference d _Z between the center of the distribution and the threshold th _Z is smaller than that in FIG.
Will be greater than. In some cases, the center of the distribution may fall below the threshold th _Z. So false negat
The probability p _{FN of} ive is also between FIG. 42 and FIG. 43, and the reliability of the detection result is also between FIG. 42 and FIG. If it is necessary to reduce the variance, as in the processing example of FIG.
It is preferable to increase the number of samples. For example, in the case of a moving image, a detection value obtained from another frame is used.

42 to 45, it is assumed that the bit information embedded in the digital watermark = 1, but in the case of bit information = 0, the sign of the standardized value Z is only inverted, and The correspondence of the standardized value Z to the variance of the detected values is shown in FIG.
2 to 45. Further, the standardized value Z when the digital watermark is not embedded is distributed around the origin.

[Detection Process Using Standardization] In a digital watermark embedding detection method that expresses bit information by a combination of two digital watermark patterns, a digital watermark detection process procedure in which a method that uses standardization of detection values is applied to the detection process. This will be described with reference to the processing flows of FIGS. 46 and 47. In the processing flow of FIG. 46, when the detected values of two digital watermarks are converted into the t statistic by [Equation 30] or [Equation 32] and then compared with the threshold value, the applied t statistic is represented by [Equation 30]. It is an example of a process in which which one of [Equation 32] is used is determined in advance. On the other hand, in the processing flow of FIG.
2] is an example of a process of performing a process of determining which of 2) is to be performed based on the variance of the detection values of the two digital watermarks.

First, each step of the processing flow of FIG. 46 will be described. In step S4601, an image that is the target of digital watermark detection processing is input. Then step S4
At 602, the digital watermark pattern 1 and the digital watermark pattern 2 are detected in the input image. Although this flow describes an example in which two digital watermark patterns are applied, when three or more digital watermark patterns are applied, detection processing of each of those digital watermark patterns is executed. In the detection process, a plurality of detection values are acquired.

In step S4603, variances are independently obtained from a plurality of detection values of the digital watermark 1 and the digital watermark 2, and the above-mentioned [Equation 29] or [Equation 31] is calculated according to a predetermined t-statistic calculation method. Is applied to obtain the variance of the detected values, and then either [Equation 30] or [Equation 32] is applied to calculate the t statistic. Step S4
In 604, the calculated t statistic is compared with a threshold value based on the judgment formula of [Equation 28], and step S4 is performed.
At 605, the presence or absence of a digital watermark and the bit information as embedded information are determined based on the comparison result. In step S4606, results such as the presence / absence of a digital watermark and embedded information based on the determination result are output, and the process ends.

Next, each step of the processing flow of FIG. 47 will be described. In step S4701, an image to be a digital watermark detection processing target is input. Then step S4
At 702, the digital watermark pattern 1 and the digital watermark pattern 2 are detected in the input image. Although this flow describes an example in which two digital watermark patterns are applied, when three or more digital watermark patterns are applied, detection processing of each of those digital watermark patterns is executed. In the detection process, a plurality of detection values are acquired.

In step S4703, variances are independently obtained from a plurality of detection values of the digital watermark 1 and the digital watermark 2, and it is verified whether it is possible to determine that the mother variances of the detection values of the two digital watermarks are equal. That is, for each of the digital watermark pattern 1 and the digital watermark pattern 2, a predetermined value n ₁ or n ₂ or more of detection value data is acquired, and the detection values of the digital watermark pattern 1 and the digital watermark pattern 2 are dispersed. Of the digital watermark pattern 1 and the digital variance of the digital watermark pattern 2 are equal to each other. When it is determined that they are equal, the above-mentioned [Equation 30] is used, and when it is determined that they are not equal, [Equation 32] is used to obtain the t statistic.

If it is determined that the population variances are equal, then in step S4704, the variance of the detected values is obtained by applying the above [Equation 29], and further, the [T] statistic is calculated by applying the [Equation 30]. . If it is determined that the population variances are not equal, then in step S4704, the variance of the detection values is obtained by applying the above [Formula 31], and the [Formula 32] is further applied to calculate the t statistic.

In step S4705, a comparison process is performed for the calculated t statistic with a threshold value based on the judgment formula of [Equation 28]. In step S4706, the presence / absence of a digital watermark and embedded information as embedded information Determine bit information. In step S4707, results such as the presence / absence of a digital watermark and embedded information based on the determination result are output, and the process ends.

In the above embodiment, the standardization in the case of using two digital watermark patterns for expressing 1-bit information has been described above. However, the configuration of the present invention is as follows.
It is not limited to two digital watermark patterns,
In the case of three or more, Ba is used as a one-way analysis of variance for population mean and Ba as the test of variance uniformity for population variance.
The rtlett method and the Hartley method can be applied.

[System Configuration] The series of processes described in the above embodiments can be executed by hardware, software, or a composite configuration of both. When executing the processing by software, the program in which the processing sequence is recorded is installed in the memory in the data processing device incorporated in the dedicated hardware and executed, or a general-purpose computer capable of executing various kinds of processing is executed. It is possible to install and run the program. When the series of processes is performed by software, the program forming the software is installed in, for example, a general-purpose computer or a one-chip microcomputer. FIG. 48 shows
An example of a system configuration of an apparatus that executes the series of processes described above, specifically, at least one of the digital watermark embedding and the detection, is shown.

The system is a CPU (Central Processing).
Unit) 4802. CPU (Central Processing U
(nit) 4802 is for various application programs,
An OS (Operating System) is actually executed. ROM
A (Read-Only-Memory) 4803 stores a program executed by the CPU 4802 or fixed data as a calculation parameter. RAM (Random Access Memory)
Reference numeral 4804 is used as a storage area and a work area for programs executed in the processing of the CPU 4802 and parameters that change appropriately in the program processing. CPU4802, ROM4803, RAM480
4 and the hard disk 4805 are connected by a bus 4801, and data transfer can be executed mutually. Further, data transfer with various input / output devices connected to the input / output interface 4811 is possible.

The keyboard 4812 and mouse 4813 are C
It is operated by the user to input various commands to the PU 4802, and is operated by the user when inputting command input data.
14 is input.

The drive 4809 is a floppy (registered trademark) disk or CD-ROM (Compact Disc ReadOnly M).
emory), MO (Magneto optical) disc, DVD (Digi
tal Versatile Disc), a magnetic disk, a semiconductor memory, and the like, which is a drive that executes recording / reproduction of the removable recording medium 4810, and executes a program or data reproduction from each removable recording medium 4810 and a program or data storage in the removable recording medium 4810. .

The CPU 4802 has a keyboard 4812 and a mouse 481 via the input / output interface 4811.
When a command is input via 3 etc., according to the input,
A program stored in a ROM (Read Only Memory) 4803 is executed.

The image to be embedded in the digital watermark or the image to be detected in the above-described embodiment is input to the input unit 4807 such as a camera 4871 or other input device such as a data input device such as a scanner or a drive 4809. Floppy disk, CD- connected to
ROM (Compact Disc Read Only Memory), MO (Magnet
o optical) disk, DVD (Digital Versatile Dis)
c), a removable recording medium 4810 such as a magnetic disk or a semiconductor memory. Note that this system has a configuration in which voice data can also be input via the microphone 4872. Further, the data received via the communication unit 4808 can be processed as image data to be embedded in the digital watermark or image data to be detected.

The CPU 4802 is not limited to the ROM storage program, but the program stored in the hard disk 4805, the program transferred from the satellite or the network, received by the communication unit 4808 and installed in the hard disk 4805, or the drive 4809.
It is also possible to load the program read from the removable recording medium 4810 mounted on the hard disk 4805 and installed in the hard disk 4805 into the RAM (Random Access Memory) 4804 and execute the program.

In the system having the configuration shown in FIG. 48, the CPU 4802 performs the processing according to each of the above-described embodiments or the processing performed according to the above-described block diagram and flowchart. And the CPU4802
Displays the processing result as needed, for example, via an input / output interface 4811 to an LCD (Liquid Crysta
It outputs to a display device 4861 such as a lDisplay) or a CRT and a speaker 4862 via an output unit 4806. Further, the processed data can be transmitted from the communication unit 4808 and further stored in a recording medium such as the hard disk 4805.

Execution programs for various processes are stored in the hard disk 480 as a recording medium built in the system.
5 or the ROM 4803 can be recorded in advance.
Alternatively, the program is a floppy disk, CD-R
OM (Compact Disc Read Only Memory), MO (Magneto
optical) disc, DVD (Digital Versatile Disc),
It can be stored (recorded) in a removable recording medium 4810 such as a magnetic disk or a semiconductor memory temporarily or permanently. Such removable recording medium 4810 can be provided as so-called package software.

The program is installed in the computer from the removable recording medium 4810 as described above, and is also wirelessly transferred from the download site to the computer via an artificial satellite for digital satellite broadcasting or LAN (Local Area Network). ), By wire transfer to a computer via a network such as the Internet, and in the computer, the program transferred in this way can be received by the communication unit 4808 and installed in the built-in hard disk 4805.

Here, in the present specification, the processing steps for writing a program for causing a computer to perform various kinds of processing do not necessarily have to be processed in time series in the order described in the flow chart, but in parallel. Alternatively, it also includes processes that are executed individually (for example, parallel processes or processes by objects).

Further, the program may be processed by one computer or may be processed in a distributed manner by a plurality of computers. Further, the program may be transferred to a remote computer and executed.

The present invention has been described in detail above with reference to the specific embodiments. However, it is obvious that those skilled in the art can modify or substitute the embodiments without departing from the scope of the present invention. For example, in the above description, the area in which the digital watermark pattern is embedded is the image area that is the image itself, but this method can be applied not only to the image area but also to the frequency area of the image. That is, DCT (Discr
ete Cosine Transform) and DFT (Discrete Fourier Tra)
nsform) and Wavelets Transfor
m) and other signals in the frequency domain, and DF
If it is T, it can be applied to its amplitude spectrum and phase spectrum.

In the embodiment, the image signal has been described as a target of digital watermark embedding and detection processing, but the configuration of the present invention is not limited to these specific image signals, and digital watermark embedding processing is executed. The present invention can be applied to voice signals and various data signals that are likely to occur.
That is, the present invention has been disclosed in the form of exemplification, and should not be limitedly interpreted. In order to determine the gist of the present invention, the section of the claims described at the beginning should be taken into consideration.

The various processes described in the specification are
The processing may be executed not only in time series according to the description, but also in parallel or individually according to the processing capability of the device that executes the processing or the need. Further, the system in the present specification is a logical set configuration of a plurality of devices,
The devices of the respective configurations are not limited to being in the same housing.

[0272]

According to the configuration of the present invention described above, various effects described below can be obtained. First, according to the present invention, a plurality of digital watermark patterns are used to represent bit information, and bit information is acquired by detecting differences in a plurality of digital watermarks. By utilizing the difference between the averages of the digital watermark detection values, the detection values can be standardized and the detection results can be evaluated in a unified manner.

Further, according to the configuration of the present invention, a plurality of digital watermark patterns are used to represent bit information, bit information is acquired by difference detection of a plurality of digital watermarks, and a statistical amount obtained by standardization is used. Then, since it is configured to determine whether or not the digital watermark is embedded and the bit information when the digital watermark is embedded, it is false for various types of images, that is, for various images in which variations in detected values occur in various modes. It is possible to reduce the probability of positive or false negative.

Further, according to the configuration of the present invention, a plurality of digital watermark patterns are used to express bit information, bit information is acquired by difference detection of a plurality of digital watermarks, and standardization processing is performed when determining bit information. Since the t-statistic whose distribution is uniquely determined if the degree of freedom is determined is used as the statistic used in, the probability of false positive or false negative in digital watermark detection can be easily brought close to a desired value. Become.

Further, according to the configuration of the present invention, a plurality of digital watermark patterns are used to express bit information, bit information is acquired by difference detection of a plurality of digital watermarks, and standardization processing is performed when determining bit information. Since the t statistic and its distribution are variably set according to the similarity of the variances of the detected values of a plurality of digital watermarks when calculating the t statistic used in Highly accurate digital watermark information determination processing according to the population becomes possible.

[Brief description of drawings]

FIG. 1 is a diagram illustrating a relative frequency distribution of inner product values of an original image and a digital watermark pattern, and a digital watermark embedded image and a digital watermark pattern.

FIG. 2 is a diagram illustrating criteria for determining the presence or absence of a digital watermark.

FIG. 3 is a diagram illustrating a threshold value (th) applied to detection of a digital watermark.

FIG. 4 is a diagram illustrating a threshold value (th) when the probabilities p _FP and p _FN are small.

FIG. 5 is a diagram illustrating a process of embedding a plurality of digital watermark patterns in an image.

FIG. 6 is a diagram illustrating division of an original image into small areas.

FIG. 7 is a diagram illustrating allocation of the same information bit to a plurality of small areas.

FIG. 8 is a diagram for describing bit information embedding and detection processing using a digital watermark pattern.

FIG. 9 is a diagram illustrating bit information embedding and detection processing using a plurality of digital watermark patterns.

FIG. 10 is a diagram illustrating a digital watermark embedding processing device.

FIG. 11 is a diagram illustrating a process of a digital watermark detection processing device.

FIG. 12 is a diagram illustrating a bit information acquisition process by detecting a plurality of digital watermarks in a digital watermark detection processing device.

FIG. 13 is a diagram illustrating variation (dispersion) of digital watermark detection values with respect to an original image.

FIG. 14 is a diagram illustrating a division process of an original image into small regions.

FIG. 15 is a diagram illustrating variation (dispersion) of digital watermark detection values.

FIG. 16 is a diagram illustrating standardization processing for varying the magnitude (variance) of digital watermark detection values.

FIG. 17 is a diagram illustrating standardization processing in which a plurality of frames of variations (dispersions) in digital watermark detection values are set.

FIG. 18 is a diagram illustrating a problem in a bit information acquisition process by detecting a plurality of digital watermarks in a digital watermark detection processing device.

FIG. 19 is a diagram for describing problems when the variance of a plurality of digital watermark detection values is small and when the variance is large.

FIG. 20 is a diagram illustrating a problem when the variances of a plurality of digital watermark detection values are different.

FIG. 21 is a diagram illustrating an outline of digital watermark detection processing according to the present invention.

FIG. 22 is a diagram illustrating distribution of difference values of a plurality of digital watermarks.

FIG. 23 is a diagram illustrating an aspect of a method of expressing bit information using detected value differences of a plurality of digital watermarks in the configuration of the present invention.

FIG. 24 is a diagram illustrating an aspect of a method of expressing bit information using detected value differences of a plurality of digital watermarks in the configuration of the present invention.

[Fig. 25] Fig. 25 is a diagram illustrating an aspect of a method of expressing bit information using detected value differences of a plurality of digital watermarks in the configuration of the present invention.

[Fig. 26] Fig. 26 is a diagram for describing one aspect of a method of expressing bit information using detected value differences of a plurality of digital watermarks in the configuration of the present invention.

[Fig. 27] Fig. 27 is a diagram illustrating a method of expressing bit information using detected value differences of three digital watermarks in the configuration of the present invention.

[Fig. 28] Fig. 28 is a diagram for describing a method of expressing bit information using three digital watermark detection value differences in the configuration of the present invention.

FIG. 29 is a diagram showing a configuration of a digital watermark embedding processing device.

[Fig. 30] Fig. 30 is a diagram for describing one aspect of a process of allocating a digital watermark embedding strength, which can be executed in a digital watermark embedding processing device.

[Fig. 31] Fig. 31 is a diagram for describing an aspect of a digital watermark embedding strength allocation process that can be executed by a digital watermark embedding processing device.

[Fig. 32] Fig. 32 is a diagram for describing an aspect of a digital watermark embedding strength allocation process that can be executed by a digital watermark embedding processing device.

[Fig. 33] Fig. 33 is a diagram for describing an aspect of the allocation process of the digital watermark embedding strength that can be executed by the digital watermark embedding processing device.

FIG. 34 is a diagram showing a processing flow of digital watermark embedding processing.

FIG. 35 is a diagram illustrating a digital watermark embedding process based on image analysis.

FIG. 36 is a diagram showing a processing flow of digital watermark embedding processing based on image analysis of the present invention.

FIG. 37 is a diagram showing a configuration of a digital watermark detection processing apparatus of the present invention.

FIG. 38 is a diagram showing a processing flow of digital watermark detection processing of the present invention.

FIG. 39 is a diagram illustrating a standardized distribution of differences between detection values of a plurality of digital watermarks.

FIG. 40 is a diagram showing a configuration of a digital watermark detection processing apparatus of the present invention.

FIG. 41 is a diagram showing a processing flow of digital watermark detection processing of the present invention.

[Fig. 42] Fig. 42 is a diagram for describing an example of standardization processing when the variances of detected values of a plurality of digital watermarks are both small.

[Fig. 43] Fig. 43 is a diagram for describing an example of standardization processing when the variances of detected values of a plurality of digital watermarks are both large.

[Fig. 44] Fig. 44 is a diagram for describing an example of standardization processing when combinations of a plurality of detected values of a plurality of digital watermarks having a large variance and a small variance are combined.

[Fig. 45] Fig. 45 is a diagram for describing an example of standardization processing when combinations of small and large variances of detection values of a plurality of digital watermarks are combined.

FIG. 46 is a diagram showing a processing flow of digital watermark detection processing of the present invention.

FIG. 47 is a diagram showing a processing flow of digital watermark detection processing of the present invention.

[Fig. 48] Fig. 48 is a diagram illustrating an example of a system configuration that executes at least one of a process of embedding and detecting a digital watermark.

[Explanation of symbols]

1001 original image 1002 Embedded information 1003 digital watermark pattern generation key storage unit 1004 Digital Watermark Pattern Generation Unit 1005 Digital watermark pattern embedding unit 1006 Digital watermark embedded image 1101 Detection target image 1102 electronic watermark pattern generation key storage unit 1103 Digital Watermark Pattern Generation Unit 1104 Detection unit 1105 Detection information 2001 original image 2002 Embedded information 2003 Digital Watermark Pattern Generation Key Storage Unit 2004 Digital Watermark Pattern Generation Unit 2005 Digital Watermark Detection Unit 2006 Digital watermark pattern embedding unit 2007 Digital watermark embedded image 3001 Detection target image 3002 Digital Watermark Pattern Generation Key Storage Unit 3003 Digital Watermark Pattern Generation Unit 3004 Detector 3005 Detection information 3011 Digital Watermark Detection Value Acquisition Unit 3012 Detection value difference acquisition unit 3013 detection value difference determination unit 4001 detection target image 4002 Digital Watermark Pattern Generation Key Storage Unit 4003 Digital Watermark Pattern Generation Unit 4004 detector 4005 detection information 4011 Digital Watermark Detection Value Acquisition Unit 4012 Detection value standardization processing unit 4013 detection value difference determination unit 4021 variance comparison unit 4022 t statistic calculator 4802 CPU 4803 ROM 4804 RAM 4805 hard disk 4806 Output unit 4807 Input section 4808 Communication unit 4809 drive 4810 removable recording medium 4812 keyboard 4813 mouse 4814 keyboard mouse controller 4861 display device 4862 speaker 4871 camera 4872 microphone

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Yuki Matsumura             6-735 Kita-Shinagawa, Shinagawa-ku, Tokyo Soni             -Inside the corporation (72) Inventor Shunichi Soma             6-735 Kita-Shinagawa, Shinagawa-ku, Tokyo Soni             -Inside the corporation F term (reference) 5B057 CA12 CA16 CB12 CB16 CC01                       CE08 CH08                 5C063 AB05 CA23 CA29 CA36 DA07                       DA13 DB09                 5C076 AA14 BA06                 5C077 LL14 NP01 PP23 PP43 PP47                       PQ19 PQ20 TT06

Claims (11)

[Claims] 1. A digital watermark detection processing apparatus for executing digital watermark pattern detection processing, wherein detection processing of each of a plurality of digital watermark patterns is performed on an input image which is the target of digital watermark pattern detection processing, and each digital watermark pattern is detected. Digital watermark pattern detection value acquisition means for acquiring a plurality of detection values for each watermark pattern, and detection of each digital watermark pattern based on a plurality of detection values of each pattern acquired by the digital watermark pattern detection value acquisition means A standardization processing unit that executes standardization processing of the value difference and acquires a standardized value calculated as a result of the standardization processing; and embedded digital watermark information based on the comparison processing of the standardized value and the threshold acquired by the standardization processing unit. Digital watermark embedding pattern determining means for acquiring apparatus. 2. The standardization processing means compares a variance based on a plurality of detection values of each pattern acquired by the digital watermark pattern detection value acquisition means with each other, and based on an output of the dispersion comparison part. And a t statistic calculating unit that calculates a t statistic as a standardized value, wherein the digital watermark embedding pattern determination unit acquires embedded digital watermark information based on a comparison process between the t statistic and a threshold value. The electronic watermark detection processing device according to claim 1, wherein 3. The standardization processing means compares a variance based on a plurality of detection values of each pattern acquired by the digital watermark pattern detection value acquisition means with each other, and based on an output of the dispersion comparison part. The electronic watermark embedding pattern determination means includes a t statistic and a t statistic calculating unit that calculates a t distribution, and the digital watermark embedding pattern determination unit embeds the t statistic and a threshold value determined based on the t distribution. The digital watermark detection processing device according to claim 1, wherein the digital watermark detection processing device is configured to acquire digital watermark information. 4. The standardization processing means is configured to apply a predetermined t statistic calculation formula to execute a t statistic calculation process based on the output of the variance comparison unit. Item 2. The electronic watermark detection processing device according to Item 1. 5. The standardization processing means applies different t statistic calculation formulas depending on whether a plurality of variances based on detection values of a plurality of digital watermark patterns are equal or different, and the variance comparison is performed. Based on the output of the department
The electronic watermark detection processing device according to claim 1, wherein the electronic watermark detection processing device is configured to execute a statistic calculation process. 6. A digital watermark detection processing method for executing digital watermark pattern detection processing, wherein detection processing for each of a plurality of digital watermark patterns is performed on an input image which is the target of digital watermark pattern detection processing, and each digital watermark pattern is detected. A digital watermark pattern detection value acquisition step of acquiring a plurality of detection values for each watermark pattern, and detection of each digital watermark pattern based on the plurality of detection values of each pattern acquired in the digital watermark pattern detection value acquisition step Perform standardization of value differences,
A standardization processing step of acquiring a standardized value calculated as a result of the standardization processing, and a digital watermark embedding pattern determination step of acquiring embedded digital watermark information based on the comparison processing of the standardized value acquired in the standardization processing step and a threshold value. And a digital watermark detection processing method comprising: 7. The standardization processing step is based on a variance comparison step of comparing variances based on a plurality of detection values of each pattern acquired in the digital watermark pattern detection value acquisition step, and based on an output in the variance comparison step. A t-statistic calculation step of calculating a t-statistic as a standardized value, wherein the digital watermark embedding pattern determination step acquires embedded digital watermark information based on a comparison process between the t-statistic and a threshold. The electronic watermark detection processing method according to claim 6. 8. The standardization processing step is based on a variance comparison step of comparing variances based on a plurality of detection values of each pattern acquired in the digital watermark pattern detection value acquisition step, and based on an output in the variance comparison step. and a t statistic calculating step of calculating a t distribution, wherein the digital watermark embedding pattern determining step includes embedding based on a comparison process between the t statistic and a threshold determined based on the t distribution. The digital watermark detection processing method according to claim 6, wherein digital watermark information is acquired. 9. The standardization processing step applies a predetermined t statistic calculation formula to execute a t statistic calculation process based on the output of the variance comparison unit. The described digital watermark detection processing method. 10. The standardization processing step applies different t statistic calculation formulas depending on whether a plurality of variances based on detected values of a plurality of digital watermark patterns are equal or different, and the variance comparison is performed. Based on the output of the department
7. The digital watermark detection processing method according to claim 6, wherein a statistical amount calculation process is executed. 11. A program for executing a digital watermark pattern detection process on a computer system, wherein the detection process of each of a plurality of digital watermark patterns is executed on an input image which is a target of the digital watermark pattern detection process. A digital watermark pattern detection value acquisition step of acquiring a plurality of detection values for each digital watermark pattern, and a digital watermark pattern based on the plurality of detection values of each pattern acquired in the digital watermark pattern detection value acquisition step Execute the standardization process of the detection value difference of
A standardization processing step of acquiring a standardized value calculated as a result of the standardization processing, and a digital watermark embedding pattern determination step of acquiring embedded digital watermark information based on a comparison processing of the standardized value acquired in the standardization processing step and a threshold value. And a program having: