JP2001053949A - Digital watermark inserting system, digital watermark characteristic table generating system and digital watermark characteristic parameter table generating system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To insert a digital watermark by automatically controlling a digital watermark strength by calculation the resistance evaluation value of the digital watermark with a uniform reference regardless of the kind of attack in a system for inserting the digital watermark by automatically controlling the strength. SOLUTION: A detection rate converting part 13 converts detection rate data, which are outputted from a detection rate calculating part 12, corresponding to an attack parameter and the detection rate data corresponding to a picture quality after attack are found. The provided data are inputted to a resistance evaluation value calculating part 14, the resistance evaluation value is calculated by common statistic processing independent of the kind of attack and a digital watermark characteristic table is generated by a characteristic table generating part 15 together with a picture quality deterioration degree outputted from a picture quality deterioration degree calculating part 11. The provided digital watermark characteristic table is inputted to a storage 1 by an input device 5 and used for controlling the digital watermark strength.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital watermark inserting system, a digital watermark characteristic table generating system, and a digital watermark characteristic parameter table generating system, and more particularly to a digital watermark inserting system for converting information into a digital watermark and inserting it into an image, and a digital watermark inserting system. The present invention relates to a system for generating a digital watermark characteristic table and a digital watermark characteristic parameter table to be used for the electronic watermarking.

[0002]

2. Description of the Related Art Conventionally, this type of digital watermarking system has
For example, refer to the document "October 1998, 1998 Video Media Processing Symposium, 3rd Symposium, 47-
48, I-3.04 Computation of Evaluation Value of Digital Watermark against Attack and Its Application ”, the digital watermark strength is automatically determined and used for the purpose of inserting a digital watermark. ing.

[0003] When a digital watermark is inserted, the image quality is degraded, and the degree of this degradation increases as the digital watermark strength increases. On the other hand, the resistance to attack increases as the digital watermark strength increases. Thus, there is a trade-off between the image quality after digital watermark insertion and the resistance to attack. This relationship changes depending on the content and type of the image. Therefore, when inserting a digital watermark, it is necessary to calculate a digital watermark strength suitable for each image and insert a digital watermark.

If the degree of image quality deterioration due to the insertion of a digital watermark is defined as the image quality deterioration degree D, and the resistance of the digital watermark to attack is quantified as the resistance evaluation value V, the above-mentioned document describes the objective function Z Is determined as the optimum digital watermark strength.

[0005]

(Equation 1) Here, a is a parameter that adjusts the rate at which the durability evaluation value V and the image quality degradation degree D contribute to the objective function Z. More specifically, a resistance evaluation value V and a degree of image quality degradation D are calculated for each digital watermark strength, and an objective function value of Equation 1 is calculated for each digital watermark strength, and this is maximized. Digital watermark strength is required. Since the value of the robustness evaluation value V and the value of the image quality degradation D for each digital watermark strength differ depending on the type of image, the values of the tolerance evaluation value V and the image quality degradation D are calculated in advance for each type, and If it is selected and used according to the type, the digital watermark strength suitable for the input image can be calculated.

Hereinafter, a table that describes the relationship between the image quality degradation degree D and the robustness evaluation value V for each digital watermark strength for each image type (category) will be referred to as a digital watermark characteristic table. .

FIG. 20 is a block diagram showing a configuration of a digital watermark insertion system that determines an optimum digital watermark strength based on the above principle and inserts a digital watermark into an image.

In FIG. 1, a category classification unit 3 obtains a category to which an input image belongs, and outputs a category index to the storage device 1. The storage device 1 includes a category classification unit 3
From the category index output from, and the digital watermark strength output from the digital watermark strength calculation unit 4, the corresponding image quality deterioration degree and resistance evaluation value are obtained and output to the digital watermark strength calculation unit 4.

The category classifying unit 3 calculates the feature amount of the image, determines which category the input image belongs to based on the obtained feature amount, and outputs a category index indicating the category. Specifically, the category classification unit holds the values of the feature values that give the boundaries of the categories, compares them with the calculated feature values, and performs classification.
The feature amount is, for example, the activity of the entire image (the average value of the AC frequency component) or the JND (Just Not).
The average value of an image (Cableable Distortion), the number of colors used, entropy and the like can be considered.

[0010] Alternatively, if this category is a medical image, C
The type of the image such as G or animation is determined, and the type of the image may be automatically estimated from the feature amount of the image, or the user may be able to explicitly specify the category.

The digital watermark strength calculation unit 4 outputs the digital watermark strength to the storage device 1 and, as a result, the image quality degradation degree and the resistance evaluation value output from the storage device 1 and the digital watermark strength constraint input from the user. Based on the information, the optimum digital watermark strength is determined and output to the digital watermark insertion unit 2. The digital watermark insertion unit 2 converts the embedded data into a digital watermark, inserts a digital watermark into the input image with the optimum digital watermark strength output from the digital watermark strength calculation unit 4, and outputs a digital watermark inserted image. The storage device 1 includes:
Digital watermark characteristic table generation system 12 via input device 5
10 are connected.

The input device 5 is a device that receives digital watermark characteristic table data output from the digital watermark characteristic table generation system and outputs the data to the storage device. That is, the digital watermark characteristic table data received from the digital watermark characteristic table generation system is received and stored in the storage device 1.

The digital watermark characteristic table data is temporarily recorded on a recording medium such as a floppy disk, CD-ROM, or magneto-optical disk. The input device 5 may be a device that reads these recording media. Still further, the recording medium may be a storage device such as a hard disk provided in a server device or the like, and the input device may be a device that reads data from the storage device via a network.

Digital watermark characteristic table generation system 1210
Is composed of an image quality deterioration degree calculation unit 11, a detection rate calculation unit 12, a tolerance evaluation value calculation unit 1211, and a characteristic table generation unit 1212. The image quality deterioration degree calculation unit 11 calculates the image quality deterioration degree and outputs it to the characteristic table generation unit 1212. The detection rate calculation unit 12 calculates the detection rate and outputs it to the resistance evaluation value calculation unit 1211. The resistance evaluation value calculation unit 1211 includes the detection rate calculation unit 1
Then, the resistance evaluation value is calculated from the detection rate output from No. 2 and output to the characteristic table generation unit 1212. Characteristic table generator 1212
Generates a digital watermark characteristic table from the image quality deterioration degree output from the image quality deterioration degree calculation unit 11 and the robustness evaluation value output from the robustness evaluation value calculation unit 1211, and outputs it to the input device 5.

Next, the operation of the system shown in FIG. 20 will be described.

The input image is input to the category classification unit 3. The category classification unit 3 calculates the feature amount of the image, determines to which category the input image belongs based on the obtained feature amount, and outputs a category index indicating the category. The category index output from the category classification unit 3 is input to the storage device 1.

The storage device 1 stores a digital watermark characteristic table. In the storage device 1, the digital watermark strength calculation unit 4
Then, based on the digital watermark strength input from the above and the category index output from the category classification unit 3, the corresponding image quality deterioration degree and resistance evaluation value are output to the digital watermark strength calculation unit 4.

In the digital watermark strength calculating section 4, after the input image is input to the system and the category index is calculated in the category classifying section 3, the optimum digital watermark strength is calculated based on the digital watermark strength constraint information input by the user. Is calculated. Here, the digital watermark strength that maximizes the objective function shown in Expression 1 is selected. That is, the objective function value of Expression 1 is calculated from the image quality degradation degree and the resistance evaluation value output from the storage device 1 corresponding to each digital watermark intensity output to the storage device 1, and the digital watermark that maximizes the objective function value is calculated. Find the strength. The digital watermark strength constraint information input by the user is a lower limit value of the image quality of the digital watermark inserted image, a lower limit value of the resistance evaluation value, or a parameter a of Expression 1, which is also a digital value. Used to determine watermark insertion strength.

The optimum digital watermark strength output from the digital watermark strength calculation section 4 is input to the digital watermark insertion section 2. The digital watermark insertion unit 2 converts the input embedded data into a digital watermark signal,
The digital watermark is inserted into the input image at the optimum digital watermark strength output from. Then, the obtained image is output as a digital watermark inserted image.

As the digital watermark insertion algorithm, any algorithm can be used as long as the user can specify the digital watermark strength or the like as a parameter at the time of digital watermark insertion. For example, Japanese Unexamined Patent Publication No. 9-191394 or the document "December 1997, IEE Transactions on Image Processing, Vol. IP-6, No. 12, pp. 1673-1687 (IEEE TRANSACTIONS) ONIMAG
E PROCESSING, VOL. IP-6, NO.
12, 1997) ").

On the other hand, the digital watermark characteristic table stored in the storage device 1 is generated by the digital watermark characteristic table generation system 1210 and input to the storage device 1 via the input device 5.
Next, the operation of the digital watermark characteristic table generation system 1210 will be described.

The image quality deterioration degree calculating section 11 inserts a digital watermark into a plurality of images classified in advance by changing the digital watermark strength, and compares the obtained image with the image before the digital watermark is inserted. Thus, the deterioration amount of the image quality is calculated. Then, the obtained image quality deterioration amounts are totaled for each category index and digital watermark strength, and the image quality deterioration degree is calculated.

Here, the operation of the image quality deterioration degree calculating section 11 will be described with reference to FIG. FIG. 5 is a flowchart showing the operation of the image quality deterioration degree calculation unit 11. In the figure, the image quality degradation degree calculation unit 11 first inserts a digital watermark into a plurality of images classified in advance by changing the digital watermark strength (step S111). Next, the image obtained by the digital watermark insertion is compared with the image before the digital watermark is inserted, and the deterioration amount of the image quality is calculated (step S112). Then, the calculated image quality deterioration amount is totaled for each category index and digital watermark strength, and
The degree of image quality deterioration is calculated (step S113).

The detection rate calculation unit 12 inserts a digital watermark into a plurality of images classified in advance by changing the digital watermark strength, and performs image processing, image encoding, and the like on the obtained image. Perform any possible attacks. The attack is executed by changing an attack parameter which is a parameter for adjusting the degree of the attack. Next, detection of a digital watermark is attempted on an image obtained for each attack parameter, and a detection result indicating whether or not the digital watermark has been detected or how much has been detected is obtained and quantified. Then, the detection result of the digital watermark is
The detection rate is calculated by totaling for each category index, digital watermark strength, and attack parameter.

Here, the operation of the detection rate calculation section 12 will be described with reference to FIG. FIG.
6 is a flowchart showing the operation of the embodiment. In the figure, the detection rate calculation unit 12 first inserts a digital watermark into a plurality of images classified in advance by changing the digital watermark strength (step S121).

Next, an attack that can occur in image processing, image encoding, or the like is performed on the image obtained by inserting the digital watermark (step S122). This attack is executed by changing an attack parameter which is a parameter for adjusting the degree of attack.

Further, detection of a digital watermark is attempted on an image obtained for each attack parameter, and a detection result indicating whether or not the digital watermark has been detected or how much has been detected is obtained and quantified. (Step S123). Then, the detection results of the digital watermarks are totaled for each category index, digital watermark strength, and attack parameter, and the detection rate is calculated (step S12).
4).

The detection rate output from the detection rate calculation section 12 is input to the tolerance evaluation value calculation section 1211. The resistance evaluation value calculation unit 1211 considers the detection rate as a function of an attack parameter, and calculates a resistance evaluation value by performing statistical processing on the function.

Several methods are conceivable for calculating the resistance evaluation value from the relationship between the attack parameter and the detection rate.
When calculating by weighted addition according to the above document,

[0030]

(Equation 2) To calculate the resistance evaluation value v. Here, r (x) is a digital watermark detection rate when the attack parameter is x, and w (x) is a weighting function. Note that statistical processing other than the weighted average may be performed.

The image quality deterioration degree output from the image quality deterioration degree calculation unit 11 and the resistance evaluation value output from the resistance evaluation value calculation unit are input to the characteristic table generation unit 1212. The characteristic table generation unit 1212 generates a digital watermark characteristic table by collecting the image quality deterioration degree and the resistance evaluation value for each category index and digital watermark strength. The obtained digital watermark characteristic table is output to the input device 5. The digital watermark characteristic table input to the input device 5 is stored in the storage device 1.

[0032]

The above-described conventional digital watermark insertion system can automatically determine the digital watermark strength according to an image and insert a digital watermark. However, the above-mentioned conventional system has the following problems.

The first problem is that it is necessary to change the method of calculating the resistance evaluation value for each attack. This is because the attack parameters are different for each attack.

If the attack is JPEG compression, the attack parameter is a parameter for controlling image quality. If the attack is noise added, the standard deviation of noise and the SNR (Signal to Noise R) of the image after noise addition are used.
atio, signal-to-noise ratio). As described above, since the attack parameter differs depending on the attack, it is necessary to appropriately set the resistance evaluation value calculation parameter for each attack. For example, when calculating a resistance evaluation value by a weighted average, it is necessary to design a weighting function to be used for each attack. Further, when the user customizes the process of calculating the resistance evaluation value, the user needs to adjust the parameters for calculating the resistance evaluation value for each attack, which increases the burden on the user.

A second problem is that if the detection rate is statistically processed with respect to the attack parameter to calculate the resistance evaluation value, there is a possibility that an appropriate resistance evaluation value cannot be calculated. This is because the correlation between the attack parameter and the image quality of the image obtained after the attack is not always high.

It is considered that the value of the image after the attack changes depending on the image quality of the image after the attack, and the importance of the detection result of the watermark changes accordingly. Therefore,
It is desirable that a resistance evaluation value calculation parameter such as a weighting function value can be controlled depending on the image quality after the attack. However, the correlation between the attack parameter and the image quality is not always high, and it is difficult to appropriately reflect the evaluation on the image quality after the attack on the tolerance evaluation value.

The present invention has been made to solve the above-mentioned drawbacks of the prior art, and its purpose is to calculate the digital watermark resistance evaluation value according to a unified standard regardless of the type of attack, and to improve the image quality after the attack. Watermark insertion system that automatically adjusts the digital watermark strength using a resistance evaluation value appropriately reflecting the evaluation of the digital watermark and inserts the digital watermark, and a digital watermark characteristic table used for controlling the digital watermark strength in the digital watermark system And a system for generating a digital watermark characteristic parameter table.

[0038]

SUMMARY OF THE INVENTION A digital watermark insertion system according to the present invention is a system for automatically adjusting the digital watermark strength according to an image and inserting a digital watermark, and calculating a feature amount of an input image. Category classification means for classifying and outputting category index data, inserting a digital watermark by changing the digital watermark strength for a plurality of images, and performing image processing and image compression on the obtained images. By changing the attack parameter, which is a parameter for adjusting the degree of possible attack, performing an attack, detecting an electronic watermark on the obtained image, totaling the obtained detection results, and setting a predetermined value. Category index value, a predetermined digital watermark intensity value, a digital watermark detection rate is calculated for each predetermined attack parameter value, A detection rate calculating means for outputting as a detection rate data, in the first detection rate data,
Category index, from a series of detection rate data where the digital watermark strength is the same value, each of the predetermined attacked image quality evaluation values representing the image quality of the image after the attack,
Using information describing the relationship with the attack parameters,
The detection rate conversion for calculating the detection rate for each of the predetermined attacked image quality evaluation values is performed, and the detection rate is calculated for each of the predetermined category index value and the predetermined digital watermark strength value. A detection rate conversion means for calculating data and outputting the data as second detection rate data; and detecting a series of detection rate data having the same category index and digital watermark strength in the second detection rate data. A resistance evaluation curve that indicates a change in the detection rate with respect to the attack image quality evaluation value and performs a statistical process on the detection rate curve to obtain a resistance evaluation value that is a value that indicates the resistance of the digital watermark to an attack. Calculating a resistance value for each of the predetermined category index value and the predetermined digital watermark strength value. A resistance evaluation value calculating unit that outputs the second resistance evaluation value data, inserts a digital watermark by changing the digital watermark strength for a plurality of images, and compares the obtained image with the image before watermark insertion. Calculating a degree of image quality deterioration representing the degree of image quality deterioration of the digital watermark inserted image with respect to each of the predetermined category index value and the predetermined digital watermark strength value; Image quality deterioration degree calculating means for outputting as data,
The second robustness evaluation value data and the second image quality degradation degree data are combined for each of the predetermined category index value and the predetermined digital watermark strength value, and the image quality degradation degree, the robustness evaluation value, and the category index are combined. Is a table describing the relationship between the digital watermark strength and the digital watermark strength table. The digital watermark property table is a table for obtaining a corresponding resistance evaluation value and image quality deterioration degree from the category index value and the digital watermark strength value. A characteristic table generating unit, storing the digital watermark characteristic table, obtaining a corresponding image quality deterioration degree and a resistance evaluation value from the category index data and the inputted digital watermark strength data, And the first
Storage means for outputting the predetermined digital watermark strength value as the digital watermark strength data to the storage means, and the first image quality output from the storage means as a result. From the degradation degree data and the first robustness evaluation value data, an objective function configured using the image quality degradation degree and the robustness evaluation value is maximized or minimized based on the input digital watermark strength constraint information. Digital watermark strength calculating means for obtaining a digital watermark strength to be converted and outputting the digital watermark strength value as an optimum digital watermark strength value, and converting the input embedded data into a digital watermark signal, and applying the optimum digital watermark strength value as a parameter to the input image. Digital watermark insertion means for inserting a digital watermark signal and generating and outputting a digital watermark inserted image.

Another digital watermark insertion system of the present invention comprises:
A system for automatically adjusting the digital watermark strength according to an image and inserting a digital watermark, comprising: a category classifying unit that calculates a feature amount of an input image, classifies the image into categories, and outputs category index data; By changing the digital watermark strength for the image and inserting a digital watermark, for the obtained image, changing the attack parameter, which is a parameter for adjusting the degree of attack that can be caused by image processing or image compression, Executes an attack, detects a digital watermark on the obtained image, totals the obtained detection results, and sets a predetermined category index value, a predetermined digital watermark strength value, and a predetermined digital watermark strength value. A detection rate calculating means for calculating a detection rate for each attack parameter value and outputting it as first detection rate data; From a series of detection rate data in which the category index and the digital watermark strength are the same value, information describing the relationship between each of the predetermined attacked image quality evaluation values representing the image quality of the image after the attack and the attack parameter is obtained. Using, by performing detection rate conversion to calculate the detection rate for each of the predetermined attacked image quality evaluation value, the predetermined category index value, for each of the predetermined digital watermark strength value Detection rate conversion means for calculating detection rate data and outputting it as second detection rate data; inserting a digital watermark by changing the digital watermark strength for a plurality of images; The degree of image quality degradation, which represents the degree of image quality degradation of the digital watermarked image, is compared with the predetermined category index value and the predetermined category index value. Image quality deterioration degree calculating means for calculating each of the determined digital watermark strength values and outputting the same as second image quality deterioration degree data; and a series of the second detection rate data having the same category index. Calculating a detection rate description parameter that is a parameter describing a detection rate surface representing a change in the detection rate with respect to the attacked image quality evaluation value and the digital watermark strength from the detection rate data; Image quality deterioration that obtains an image quality deterioration degree description parameter that is a parameter that describes an image quality deterioration degree curve representing a change in image quality deterioration degree with respect to digital watermark strength from a series of image quality deterioration degree data having the same category index in the deterioration degree data. Calculation parameter, and a detection rate description parameter is calculated for each of the predetermined category index values. Digital watermark characteristic extraction means for calculating data and image quality degradation degree description parameters, and outputting the digital watermark characteristic parameter data as digital watermark characteristic parameter data; and combining the digital watermark characteristic parameter data for each of the predetermined category index values. It is a table describing the relationship between the degradation degree description parameter and the detection rate description parameter,
A characteristic parameter table generating means for generating and outputting a digital watermark characteristic parameter table, which is a table for obtaining an image quality deterioration degree description parameter and a detection rate description parameter from the category index value, and storing the digital watermark characteristic parameter table; For the category index data, the corresponding image quality deterioration degree description parameter and detection rate description parameter are obtained, and the image quality deterioration degree curve is obtained from the image quality deterioration degree description parameter for the category index data. A first image quality deterioration degree data, which is an image quality deterioration degree, is calculated, and a detection rate indicating a relationship between the attacked image quality evaluation value and the detection rate for the digital watermark strength data is calculated from the detection rate description parameter for the category index data. Find the curve and enter Based on the evaluation value calculation parameters,
Statistical processing is performed on the detection rate curve to calculate first robustness evaluation value data which is a robustness evaluation value for the digital watermark strength data, and the first image quality degradation degree data and the first robustness evaluation data are calculated. Digital watermark characteristic calculating means for outputting value calculation data, and outputting the predetermined digital watermark strength value to the digital watermark characteristic calculating means as the digital watermark strength data, and as a result, Based on the output first image quality degradation degree data and the first robustness evaluation value data, based on the input digital watermark strength constraint information, the first image quality degradation degree and the robustness evaluation value are used. A digital watermark strength calculating means for obtaining a digital watermark strength for maximizing or minimizing the objective function and outputting it as an optimum digital watermark strength value; Digital watermark insertion means for converting the digital watermark signal into a child watermark signal, inserting the digital watermark signal into the input image with the optimal digital watermark strength value as a parameter, generating and outputting a digital watermark inserted image. I do.

The digital watermark characteristic table generation system of the present invention automatically adjusts the digital watermark strength according to an image and inserts the digital watermark into a digital watermark characteristic table used in calculating the digital watermark strength in a system for inserting a digital watermark. A system that generates a digital watermark by changing the digital watermark strength for a plurality of images and inserts the digital watermark into the obtained images.
By changing the attack parameter, which is a parameter for adjusting the degree of attack that can be caused by image processing or image compression, performing an attack that can be caused by image processing or image compression, and detecting a digital watermark on the obtained image Calculating the detection rate of the digital watermark for each of a predetermined category index value, a predetermined digital watermark strength value, and a predetermined attack parameter value, Detection rate calculation means for outputting as rate data, and in the first detection rate data,
Category index, from a series of detection rate data where the digital watermark strength is the same value, each of the predetermined attacked image quality evaluation values representing the image quality of the image after the attack,
Using information describing the relationship with the attack parameters,
The detection rate conversion for calculating the detection rate for each of the predetermined attacked image quality evaluation values is performed, and the detection rate is calculated for each of the predetermined category index value and the predetermined digital watermark strength value. A detection rate conversion unit for calculating data and outputting the data as second detection rate data; and an attack target from a series of detection rate data having the same category index and digital watermark strength in the second detection rate data. A detection rate curve representing a change in the detection rate with respect to the image quality evaluation value is calculated, and a statistical processing is performed on the detection rate curve to obtain a resistance evaluation value which is a value indicating the resistance to the attack of the digital watermark. Performing a calculation, calculating a resistance evaluation value for each of the predetermined category index value and the predetermined digital watermark strength value. A resistance evaluation value calculation unit that outputs as second resistance evaluation value data, and inserts a digital watermark by changing a digital watermark strength for a plurality of images, and compares the obtained image with an image before watermark insertion. Calculating the image quality degradation degree representing the degree of image quality degradation of the digital watermark inserted image for each of the predetermined category index value and the predetermined digital watermark strength value, Image quality deterioration degree calculating means for outputting as
The second resistance evaluation value data and the second image quality deterioration degree data are combined for each of the predetermined category index values and the predetermined digital watermark strength values, and a digital watermark characteristic table is created and output. Characteristic table generating means.

The detection rate conversion means includes: an attack parameter selection probability information which is information describing a density function of a probability that an attack parameter is selected; and an attack image quality evaluation value under a condition that the attack parameter is constant. And image quality evaluation value frequency distribution information, which is information describing the frequency distribution of the attack parameter selection probability information and the image quality evaluation value frequency distribution information. An attack parameter which is information describing a frequency distribution of attack parameters under the condition that the attack image quality evaluation value is constant based on the ratio of the connection probability to the occurrence probability of the attack image quality evaluation value. Attack parameter frequency distribution calculating means for calculating and outputting frequency distribution information; An attack parameter frequency distribution for each of the predetermined attacked image quality evaluation values is obtained from the meter frequency distribution information, and a series of detections in which the category index and the digital watermark strength are the same in the first detection rate data. From the rate data, a detection rate transition function representing a change in the detection rate with respect to the attack parameter is obtained,
A detection rate for calculating a detection rate for each of the predetermined attacked image quality evaluation values by integrating a value based on a product of the attack parameter frequency distribution and the detection rate transition function with respect to an attack parameter. Conversion means for calculating detection rate data for each of the predetermined category index value and the predetermined digital watermark strength value, and outputting the data as second detection rate data And

Further, the detection rate conversion means has image quality evaluation value attack parameter conversion information describing an image quality evaluation value attack parameter conversion function which is a function for converting the image quality evaluation value of an attacked image into an attack parameter.
Calculating the image quality evaluation value attack parameter conversion function from the image quality evaluation value attack parameter conversion information, calculating an attack parameter value corresponding to each of the predetermined attacked image quality evaluation values to obtain an equivalent attack parameter value, Among the detection rate data, the category index, from a series of detection rate data having the same digital watermark strength, a detection rate transition function representing a change in the detection rate with respect to the attack parameter is obtained, and the detection rate transition function is used. The detection rate for each of the equivalent attack parameter values, the detection rate conversion to calculate as the detection rate of the corresponding predetermined attack image quality evaluation value, the predetermined category index value, the predetermined category index value The detection rate data is calculated for each of the digital watermark strength values, and the second detection rate data is calculated. It may have a detection rate conversion value calculating means for outputting as data.

The digital watermark characteristic parameter table generation system of the present invention automatically adjusts the digital watermark strength according to an image and inserts a digital watermark into a digital watermark characteristic parameter used in calculating the digital watermark strength. A system for generating a table, in which digital watermarks are inserted into a plurality of images by changing the digital watermark strength, and the degree of an attack that can be caused by image processing or image compression is adjusted for the obtained images. By changing an attack parameter which is a parameter, performing an attack that can be caused by image processing or image compression, detecting an electronic watermark on the obtained image, totaling the obtained detection results, Category index value, a predetermined digital watermark intensity value, and a digital watermark for each predetermined attack parameter value. A detection rate calculating unit that calculates a detection rate and outputs the detection rate data as first detection rate data; and, in the first detection rate data, a category index and a series of detection rate data having the same digital watermark strength. The detection rate for each of the predetermined attacked image quality evaluation values is determined using information describing the relationship between each of the predetermined attacked image quality evaluation values representing the image quality of the image after the attack and the attack parameter. The detection rate conversion is performed to calculate the detection rate data for each of the predetermined category index value and the predetermined digital watermark strength value, and the detection rate data is output as the second detection rate data. A rate conversion unit, which inserts a digital watermark by changing the digital watermark strength for a plurality of images, compares the obtained image with the image before watermark insertion, An image quality deterioration degree representing the degree of image quality deterioration of the inserted image is calculated for each of the predetermined category index value and the predetermined digital watermark strength value, and is output as second image quality deterioration degree data. An image quality degradation degree calculating means, and a detection rate representing a change in the detection rate with respect to the attacked image quality evaluation value and the digital watermark strength from a series of detection rate data having the same category index in the second detection rate data. From the detection rate description parameter calculation for obtaining the detection rate description parameter which is a parameter describing a curved surface, and a series of image quality deterioration degree data having the same category index among the second image quality deterioration degree data, Image quality deterioration degree calculation for obtaining an image quality deterioration degree description parameter which is a parameter for describing an image quality deterioration degree curve representing a change in deterioration degree Digital watermark characteristic extracting means for calculating a detection rate description parameter and an image quality degradation degree description parameter for each of the predetermined category index values and outputting the calculated digital watermark characteristic parameter data. And a characteristic parameter table generating means for compiling digital watermark characteristic parameter data for each of the predetermined category index values, creating and outputting a digital watermark characteristic parameter table.

The detection rate conversion means includes attack parameter selection probability information, which is information describing a density function of a probability that an attack parameter is selected, and an attack image quality evaluation value under a condition that the attack parameter is constant. And image quality evaluation value frequency distribution information, which is information describing the frequency distribution of the attack parameter selection probability information and the image quality evaluation value frequency distribution information. An attack parameter which is information describing a frequency distribution of attack parameters under the condition that the attack image quality evaluation value is constant based on the ratio of the connection probability to the occurrence probability of the attack image quality evaluation value. Attack parameter frequency distribution calculating means for calculating and outputting frequency distribution information; An attack parameter frequency distribution for each of the predetermined attacked image quality evaluation values is obtained from the meter frequency distribution information, and a series of detections in which the category index and the digital watermark strength are the same in the first detection rate data. From the rate data, a detection rate transition function representing a change in the detection rate with respect to the attack parameter is obtained,
A detection rate for calculating a detection rate for each of the predetermined attacked image quality evaluation values by integrating a value based on a product of the attack parameter frequency distribution and the detection rate transition function with respect to an attack parameter. Conversion means for calculating detection rate data for each of the predetermined category index value and the predetermined digital watermark strength value, and outputting the data as second detection rate data And the following.

The detection rate conversion means has image quality evaluation value attack parameter conversion information describing an image quality evaluation value attack parameter conversion function which is a function for converting the image quality evaluation value of the attacked image into an attack parameter. The image quality evaluation value attack parameter conversion function is obtained from the image quality evaluation value attack parameter conversion information, the attack parameter value corresponding to each of the predetermined attacked image quality evaluation values is calculated as an equivalent attack parameter value, and the first attack parameter value is calculated. Among the detection rate data, the category index, from a series of detection rate data in which the digital watermark strength is the same value, determine a detection rate transition function representing a change in the detection rate with respect to the attack parameter, using the detection rate transition function, The detection rate for each of the equivalent attack parameter values is Then, the detection rate conversion is performed to calculate the detection rate data to be calculated as the detection rate of the determined attack image quality evaluation value, and the detection rate data is calculated for each of the predetermined category index value and the predetermined digital watermark strength value. , A detection rate conversion value calculating unit that outputs the detection rate as second detection rate data.

According to the present invention, a detection rate conversion unit is provided in a digital watermark characteristic table / characteristic parameter table generation system for creating a digital watermark characteristic table / characteristic parameter table used in a digital watermark insertion system. Thus, the detection rate for the attack parameter is converted into a detection rate for the image quality evaluation value of the image after the attack. The attack parameter changes depending on the attack, but the image quality evaluation value does not depend on the attack. Therefore, it is not necessary to change the parameter of the statistical processing used for calculating the resistance evaluation value for each attack. Further, by calculating the detection rate for the image quality evaluation value after the attack, it becomes easier to reflect human subjective judgment in the resistance evaluation value.

[0047]

Next, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of a digital watermark insertion system according to the present invention. Referring to FIG. 20, the configuration of the present system is the same as that of the conventional electronic device shown in FIG. 20 except that a digital watermark characteristic table generation system 10 is provided instead of the digital watermark characteristic table generation system 1210. The configuration is the same as that of the watermark insertion system.
The digital watermark characteristic table generation system 10 is connected to the input device 5, generates a digital watermark characteristic table, and outputs it to the input device 5.

The configuration of the digital watermark characteristic table generation system 10 is the same as that of the first embodiment except that a resistance evaluation value calculation unit 14 is provided instead of the resistance evaluation value calculation unit 1211 and a detection rate conversion unit 13 is newly provided. This is the same as the digital watermark characteristic table generation system 1210 shown in FIG. The detection rate conversion unit 13 converts the detection rate (hereinafter, referred to as a first detection rate) output from the detection rate calculation unit 12, and uses the obtained detection rate (hereinafter, referred to as a second detection rate) as a resistance evaluation value. Output to the calculation unit 14. The resistance evaluation value calculation unit 14 calculates a resistance evaluation value from the second detection rate output from the detection rate conversion unit 13, and outputs the obtained resistance evaluation value to the characteristic table generation unit 15.

Next, the operation of the digital watermark insertion system of FIG. 1 will be described.

The operation of the parts other than the digital watermark characteristic table generation system 10 is the same as that of the system shown in FIG. Also,
In the digital watermark characteristic table generation system 10, the operations of the parts other than the detection rate conversion unit 13 and the resistance evaluation value calculation unit 14 are the same as those in FIG.
Digital watermark characteristic table generation system 121 shown in FIG.
0 is the same as the operation.

The first detection rate output from the detection rate calculation section 12 is input to the detection rate conversion section 13. The detection rate conversion unit 13 converts the detection rate for each attack parameter output from the detection rate calculation unit 12 and calculates the detection rate for the image quality evaluation value of the image after the attack.

Here, the image quality evaluation value is PSNR (Pea
k Signal to Noise Ratio)
And MSE (Mean Square Error),
SNR (Sign) weighted by human visual characteristics
al to Noise Ratio)
(Weighted Signal to Noise
Ratio), JND (Just)
Index based on noticeable disposition, MOS (Mean OpinionScore)
Any index may be used as long as it is an index for evaluating image quality, such as a subjective evaluation value such as. Alternatively, for an attack in which it is difficult to grasp the degree of attack as signal degradation, a value calculated based on a value measured by subjective evaluation for a value of a sense of discomfort given by an image after the attack can be used. For example, when the attack is a rotation, the degree of discomfort to the image after the attack is obtained by a subjective evaluation, and an image quality evaluation scale based on the obtained value can be constructed to obtain an image quality evaluation value.

Hereinafter, the detection rate for the attack parameter x is represented by r (x), the detection rate for the image quality evaluation value Q of the attacked image is represented by R (Q), and the attack parameter xm ( m = 1,..., M), and the image quality evaluation value for calculating the second detection rate by the detection rate conversion unit 12 is Qn.
(N = 1,..., N). Then, from the first detection rate data r (xm) (m = 1,..., M) for a certain category index value and a certain digital watermark strength, a second detection is performed using the relationship between the attack parameter and the image quality evaluation value obtained in advance. A method for calculating the detection rate data R (Qn) (n = 1,..., N) will be described.

FIG. 2 shows the first detection rate data r (xm)
(M = 1,..., M), the second detection rate data R (Q
n) is a flowchart showing a procedure for calculating (n = 1,..., N). First, the value of a variable n, which is an index of the second detection rate data, is set to 1 (step 50).
0). Next, first detection rate data r (xm) (m = 1,
, M), R (Qn) is calculated using the relationship between the attack parameter and the image quality evaluation value (step 501).
Details of this calculation method will be described later. Then, the value of the index n is compared with N (step 502), and
If it is smaller, the value of n is increased by 1 (step 5).
03) Return to step 501. If n is greater than or equal to N, the process ends.

Thus, R (Qn) (n = 1,...,
N) is obtained. The above calculation process is performed for all of the predetermined category index values and each of the digital watermark strength values, and the result is output to the robustness evaluation value calculation unit 14.

The resistance evaluation value calculation unit 14 calculates a function R representing the change of the detection rate with respect to the image quality evaluation value Q of the attacked image from the second detection rate output from the detection rate conversion unit 13.
(Q) is obtained, and statistical processing is performed on this to calculate a resistance evaluation value. For example, when calculating the tolerance evaluation value by a weighted average, the tolerance evaluation value v is calculated by Expression 3.

[0058]

(Equation 3) Here, W (Q) is a weighting function for the image quality evaluation value. In the actual tolerance evaluation value calculation, the detection rate R (Q) and the weighting function W (Q) are discretized and calculated. For example,
The non-zero section of W (Q) is discretized and Qh (h = 0, ...,
H),

[0059]

(Equation 4) Is calculated by Alternatively, as a method of discretizing and calculating the quadrature, an arbitrary method of calculating an integral value by piecewise approximating an integrand, such as a Simpson formula or a trapezoidal formula, can be used. Alternatively, a waveform signal representing W (Q) R (Q) may be output instead of the above-described calculation by discretization, and this may be passed through an integrator to obtain an integrated value.

The above-described resistance evaluation value calculation is executed for each predetermined category index value and each digital watermark strength value, and the obtained resistance evaluation value is output to the characteristic table generation unit 15.

Here, the operation of the tolerance evaluation value calculator 14 will be described with reference to FIG. FIG. 4 is a flowchart showing the operation of the tolerance evaluation value calculation unit 14. In the figure, first, a function R (Q) representing a change in the detection rate with respect to the image quality evaluation value Q of the attacked image is obtained from the second detection rate output from the detection rate conversion unit 13 (step S14).
1). Next, statistical processing such as weighted averaging is performed on the obtained function R (Q) to calculate a tolerance evaluation value (step S142).

Next, an embodiment of the detection rate conversion unit 13 shown in FIG. 1 will be described in detail with reference to the drawings. FIG. 4 is a block diagram illustrating an embodiment of the detection rate conversion unit 13. The attack parameter frequency distribution calculation unit 203 calculates and outputs an attack parameter frequency distribution from the attack parameter selection probability information and the image quality evaluation value frequency distribution information. The detection rate conversion value calculation unit 202 calculates an attack from the attack parameter frequency distribution output from the attack parameter frequency distribution calculation unit 203 and the first detection rate for each attack parameter output from the detection rate calculation unit 12 in FIG. A second detection rate for the image quality evaluation value of the subsequent image is calculated and output.

Next, the operation of the detection rate conversion unit 201 shown in FIG. 4 will be described. The first detection rate for each attack parameter output from the detection rate calculation unit 12 in FIG.
Is input to the detection rate conversion value calculation section 202 in the detection rate conversion section 201 shown in FIG. On the other hand, the attack parameter frequency distribution information for each of the image quality evaluation values Qn (n = 1,..., N) of the image after the attack, output from the attack parameter frequency distribution calculation unit 203, is also input to the detection rate conversion value calculation unit 202. Is done.

Here, the attack parameter frequency distribution is
It is a frequency distribution (probability density function) p (x | Q) of the attack parameter x under the condition that the image quality evaluation value Q is constant.
The attack parameter frequency distribution information is information for describing the attack parameter frequency distribution p (x | Q).
This is a parameter of a curve approximating the attack parameter frequency distribution, or a discrete attack parameter frequency distribution value.

The detection rate conversion value calculation unit 202 calculates the attack parameter frequency distribution p (x | Qn) for each of the image quality evaluation values Qn (n = 1,..., N) of the attacked image from the attack parameter frequency distribution information. Ask. On the other hand, the detection rate r (xm) for each attack parameter x (m =
1,..., M), a function r (x) representing a change in the detection rate with respect to the attack parameter x is obtained.

[0066]

(Equation 5) , The second detection rate R (Qn) (n = 1,..., N)
Is calculated. Actually, as in the case of Equation 3, Equation 5
Calculate the value of

FIG. 5 is a flowchart showing the above processing performed by the detection rate conversion value calculation section 202. This processing corresponds to the processing performed in step 501 of the flowchart in FIG. Here, the attack parameter is xh
(H = 0,..., H)
Is calculated.

First, a variable h representing an index of an attack parameter is set to 0 (step 550). Next, p (xh | Qn) and r (xh) are calculated (step 551). p (xh | Qn) is calculated from attack parameter frequency distribution information for Qn. r (xh)
Is calculated from the first detection rate r (xm) (m = 1,..., M). Then, the value of h is compared with H (step 55).
2) If h is smaller than H, the value of h is increased by 1 (step 553), and the process returns to step 551. If h is greater than or equal to H, the value of R (Qn) in Equation 5 is calculated by section quadrature (step 554).

Next, calculation of attack parameter frequency distribution information will be described. The attack parameter frequency distribution information input to the detection rate conversion value calculation unit 202 is calculated by the attack parameter frequency distribution calculation unit 203. For this calculation, attack parameter selection probability information and image quality evaluation value frequency distribution information, which are calculated in advance and stored in storage means such as a memory, are used.

Here, the attack parameter selection probability is
The probability that each attack parameter occurs, that is, the probability that a person who performs an attack (hereinafter, referred to as an attacker) selects each attack parameter when executing an attack.
The attack parameter selection probability information is information for describing a density function p (x) of the attack parameter selection probability. This is a parameter of a curve approximating the density function p (x) of the attack parameter selection probability, or a value obtained by discretizing the density function value of the attack parameter selection probability. The shape of the function p (x) is determined by modeling the attack parameter selection of the attacker.

On the other hand, the image quality evaluation value frequency distribution is a frequency distribution p (Q | x) of the image quality evaluation value Q of the attacked image under the condition that the attack parameter x is constant. The image quality evaluation value frequency distribution information is information for describing the image quality evaluation value frequency distribution p (Q | x). This is a parameter of a curve approximating the image quality evaluation value frequency distribution p (Q | x) or a value obtained by discretizing the image quality evaluation value frequency distribution p (Q | x). Image quality evaluation value frequency distribution p (Q |
x) is calculated by setting an attack parameter value to x, performing an attack on a large number of images, calculating an image quality evaluation value Q of the obtained images, and further obtaining a distribution of the obtained image quality evaluation values Q. it can.

Hereinafter, a method of calculating the attack parameter frequency distribution information for the image quality evaluation value Qn in the attack parameter frequency distribution calculation unit 203 will be described. Here, an attack parameter frequency distribution value is obtained for a discrete attack parameter value, and this attack parameter value is assumed to be xj (j = 1,..., J). That is, the attack parameter frequency distribution value p (xj
| Qn).

FIG. 6 is a flowchart showing the processing in the attack parameter frequency distribution calculation unit 203. First,
An occurrence probability p (Qn) of the image quality evaluation value Qn is obtained (step 601). this is,

[0074]

(Equation 6) Is calculated by Actually, the value of Equation 6 is calculated in the same manner as in the case of Equation 3.

Next, the variable j representing the index of the attack parameter is set to 1 (step 601).

Next, a joint probability p (xj, Qn) between the attack parameter xj and the image quality evaluation value Qn is calculated (step 602). This is based on the attack parameter selection probability information and the density function value p (x
j) (j = 1,..., J) is obtained, and the image quality evaluation value frequency distribution p (Q
n | xj) (j = 1,..., J)

[0077]

(Equation 7) It is calculated by: Then, an attack parameter frequency distribution value p (xj | Qn) is calculated (step 603). This is based on Bayes' theorem,

[0078]

(Equation 8) Is calculated by

Next, j is compared with J (step 60).
4) If j is smaller than J, j is increased by 1 (step 606), and the process returns to step 602. If j is greater than or equal to J, the obtained attack parameter frequency distribution value p
From (xj | Qn) (j = 1,..., J), the attack parameter frequency distribution p (x |
A curve approximating the change in Qn) is obtained, and a parameter describing this curve is calculated as attack parameter frequency distribution information (step 605), and the process ends.

If a uniform distribution is used as the density function of the attack parameter selection probability, there is no need to store attack parameter selection probability information. In this case, instead of Equation 8,

[0081]

(Equation 9) Is calculated by As described above, when the density function of the attack parameter selection probability is uniform, the amount of accumulated data and the amount of calculation can be reduced.

Next, another embodiment of the detection rate conversion unit 13 shown in FIG. 1 will be described. FIG. 7 is a block diagram showing another embodiment of the detection rate conversion unit. Based on the image quality evaluation value attack parameter conversion information stored in the storage means such as a memory, the image quality evaluation value conversion unit 222 calculates the image quality evaluation value from the first detection rate output from the detection rate calculation unit 12 in FIG. The detection rate for the value is calculated and output as the second detection rate.

Next, the operation of the detection rate conversion unit 221 shown in FIG. 7 will be described. The detection rate conversion unit 221 has storage means such as a memory, and stores therein image quality evaluation value attack parameter conversion information. The image quality evaluation value attack parameter conversion information is information for describing a function for obtaining an attack parameter from the image quality evaluation value. Briefly,
The image quality evaluation value Qn for calculating the second detection rate (n = 1,...,
N), the attack parameter (this is represented by x'n) corresponding to each of them is described as it is. Alternatively, a function representing the change of the attack parameter with respect to the image quality evaluation value is obtained, a curve approximating the function is obtained, and the parameter describing this curve is described as image quality evaluation value attack parameter conversion information.

The function representing the change of the attack parameter with respect to the image quality evaluation value is obtained, for example, by obtaining the distribution of the image quality evaluation value Q for each attack parameter x, and averaging the distribution Q (or the mode value, median value, or It can be calculated by obtaining statistics similar to those. That is, a curve that approximates the relationship between the calculated average value and the attack parameter is obtained.

FIG. 8 shows the detection rate conversion value calculating section 222 of FIG.
6 is a flowchart showing the process in step S1. This processing corresponds to the processing performed in step 501 of the flowchart in FIG.

First, an attack parameter value x'n corresponding to the image quality evaluation value Qn is obtained using the image quality evaluation value attack parameter conversion information (step 650). Next, the relationship between the attack parameter x and the detection rate r (x) is obtained from the input first detection rate r (xm) (m = 1,..., M), and each attack parameter x′n (n = , N) is calculated as a detection rate R (Qn) for the image quality evaluation value Qn (step 651).
This process is performed for each image quality evaluation value Qn (n = 1,..., N), and the obtained detection rate R (Qn) is output as a second detection rate.

The detection rate conversion unit 2 shown in FIG.
21 has a simpler configuration than the detection rate conversion unit 201 shown in FIG.

Here, the digital watermark characteristic table generation system 1
The digital watermark characteristic table obtained by the storage device 1 and stored in the storage device 1 will be described. The number of categories for classifying the input image is K, and each of the K categories is distinguished by a category index k (k = 1,..., K). The value that can be taken as the digital watermark strength when inserting a digital watermark is s (m)
(M = 1,..., M). Further, when the category index is k and the digital watermark strength is s (m), the image quality degradation degree and the resistance evaluation value against the attack are D (k,
m) and V (k, m).

The digital watermark characteristic table in this case is, for example,
As shown in FIG. That is, in the child watermark characteristic table shown in the figure, the image quality deterioration amount and the resistance evaluation value for each category index value and each digital watermark strength are shown.

Next, another embodiment of the present invention will be described.
This will be described with reference to the drawings. FIG. 10 is a block diagram showing an embodiment of the digital watermark insertion system of the present invention, in which the system of FIG. 1 is improved so that the user can freely customize the process of calculating the robustness evaluation value.

In the following, a curve representing a change in the image quality deterioration degree with respect to the digital watermark strength is called an image quality deterioration degree curve, and a parameter describing this curve is called an image quality deterioration degree description parameter. Also, the image quality evaluation value of the image after the attack,
A surface representing a change in the detection rate with respect to the digital watermark strength is called a detection rate surface, and a parameter describing the surface is called a detection rate description parameter. Then, a table in which the image quality deterioration degree description parameter and the detection rate description parameter are described for each image type is referred to as a digital watermark characteristic parameter table.

The configuration of the system shown in FIG. 10 is such that a digital watermark characteristic calculating section 50 is provided instead of the storage device 1 shown in FIG. 1, and a digital watermark characteristic parameter table is provided instead of the digital watermark characteristic table generating system 10 shown in FIG. The point that a generation system 70 is provided is different from the system of FIG. 1, and the rest is the same as the configuration of the system of FIG.

The digital watermark characteristic calculating section 50 uses the digital watermark characteristic parameter table stored therein based on the category index output from the category classification section 3 and the resistance evaluation value calculation parameter input by the user.
The image quality degradation degree and the resistance evaluation value corresponding to the digital watermark strength output from the digital watermark strength calculation section 4 are calculated and output to the digital watermark strength calculation section 4. Further, a digital watermark characteristic parameter table generation system 70 is connected to the digital watermark characteristic calculation unit 50 via the input device 5.

The digital watermark characteristic parameter table generation system 70 includes an image quality deterioration degree calculation section 11 and a detection rate calculation section 12.
, Detection rate conversion unit 13, digital watermark characteristic extraction unit 71
And a characteristic parameter table generation unit 75.

The image quality deterioration degree calculating section 11 calculates the image quality deterioration degree and outputs it to the digital watermark characteristic extracting section 71. The detection rate calculation unit 12 calculates the first detection rate and outputs the first detection rate to the detection rate conversion unit 13. The detection rate conversion unit 13 converts the first detection rate output from the detection rate calculation unit 12 to obtain a second detection rate, and outputs the second detection rate to the digital watermark characteristic extraction unit 71. The digital watermark characteristic extraction unit 71 calculates a digital watermark characteristic parameter from the second detection rate output from the detection rate conversion unit 13 and the image quality deterioration degree output from the image quality deterioration degree calculation unit 11, and calculates a characteristic parameter table. Output to the generation unit 75.

Here, the operation of the digital watermark characteristic extraction unit 71 will be described with reference to FIG. FIG. 14 is a flowchart showing the operation of the digital watermark characteristic extraction unit 71.
In the figure, first, a parameter of a curve approximating an image quality deterioration degree curve is calculated as an image quality deterioration degree description parameter for each category index (step S71).
1). On the other hand, for each category index, a parameter of a curved surface that approximates the relationship between the image quality evaluation value of the attacked image, the digital watermark strength, and the detection rate is calculated as a detection rate description parameter (step S712). In addition,
The order in which step S711 and step S712 are performed may be reversed.

The characteristic parameter table generating section 75 generates a digital watermark characteristic parameter table from the digital watermark characteristic parameters output from the digital watermark characteristic extracting section 71 and outputs the table to the input device 5.

Next, the operation of the system shown in FIG. 10 will be described. The operation of the other parts except the digital watermark characteristic calculation unit 50 and the digital watermark characteristic parameter table generation system 70 is the same as that of the digital watermark insertion system shown in FIG.

The digital watermark characteristic calculating section 50 has a storage means such as a memory, and stores therein a digital watermark characteristic parameter table. Then, based on the category index output from the category classification unit 3, an image quality deterioration degree description parameter and a detection rate description parameter are obtained. An image quality deterioration degree curve representing the relationship between the digital watermark strength and the image quality deterioration degree is obtained from the image quality deterioration degree description parameter, and the image quality deterioration degree corresponding to the digital watermark strength output from the digital watermark strength calculation unit 4 is obtained.

On the other hand, from the detection rate description parameter, a function R representing the change of the detection rate with respect to the image quality evaluation value Q in the digital watermark strength output from the digital watermark strength calculation section 4 is shown.
(Q) is obtained, and statistical processing is performed on this to calculate a resistance evaluation value. The statistical processing performed here is the same as the processing in the tolerance evaluation value calculation unit 14 in FIG.

However, the parameters used by the user for calculating the resistance evaluation value can be controlled by the resistance evaluation value calculation parameter. For example, when the tolerance evaluation value is calculated by the weighted average shown in Expression 3, the weighting function w (Q) is a resistance evaluation value calculation parameter,
By appropriately setting this in advance, the user can customize the process of calculating the resistance evaluation value according to the application. Then, the obtained image quality deterioration degree and resistance evaluation value are output to the digital watermark strength calculation unit 4.

The digital watermark characteristic parameter table generation system 70 generates a digital watermark characteristic parameter table to be stored in storage means such as a memory in the digital watermark characteristic calculation section 50.

In the digital watermark characteristic parameter table generation system 70, the image quality deterioration degree calculation unit 11 and the detection rate calculation unit 12
Performs the same operation as that of the digital watermark characteristic table generation system 10 of FIG. The image quality deterioration degree output from the image quality deterioration degree calculation unit 11 is input to the digital watermark characteristic extraction unit 71, and the first detection rate output from the detection rate calculation unit 12 is:
It is input to the detection rate conversion unit 13. The operation of the detection rate conversion unit 13 is the same as the operation of the detection rate conversion unit 13 in the digital watermark characteristic table generation system 10 of FIG.

The second detection rate output from detection rate conversion section 13 is input to digital watermark characteristic extraction section 71. The digital watermark characteristic extraction unit 71 calculates a parameter of a curve approximating the image quality deterioration degree curve as an image quality deterioration degree description parameter for each category index. On the other hand, for each category index, a parameter of a curved surface approximating the relationship between the image quality evaluation value of the attacked image, the digital watermark strength, and the detection rate is calculated as a detection rate description parameter. Then, the obtained image quality deterioration degree description parameter and detection rate description parameter are stored in the characteristic parameter table generation unit 7.
5 is output. The characteristic parameter table generation unit 75 generates an electronic watermark characteristic parameter table by combining the image quality deterioration degree description parameter and the detection rate description parameter for each category index. The obtained digital watermark characteristic parameter table is output to the input device 5.

The digital watermark characteristic parameter table input to the input device 5 is stored in storage means such as a memory in the digital watermark characteristic calculating section 50.

In another embodiment of the present invention shown in FIG. 10, the user can freely customize the process of calculating the tolerance evaluation value according to the application. Further, the resistance evaluation value can be calculated using a common resistance evaluation value calculation parameter regardless of the attack. Therefore, there is no need for the user to customize the tolerance evaluation value calculation process individually for each attack.

[0107]

Next, a detection rate conversion unit 20 shown in FIG.
A first embodiment will be described. Here, an example in which second detection rate data is calculated for a certain category index value and a certain digital watermark intensity value, and an example in which a detection rate description parameter is further calculated from the obtained second detection rate data will be described. . Considering JPEG compression as an attack, the attack parameter x is a JPEG image quality parameter taking a value between 1 and 100. In the case of JPEG, the compression ratio and the image quality are adjusted by controlling the values of the Huffman table and the quantization table, and the image quality parameter is a parameter used for controlling these values. At this time, the attack parameters for calculating the first detection rate are x1 = 1, x2 = 35, x3 = 60, x4 = 8
0, x5 = 90, x6 = 94, x7 = 96, x8 = 9
8, x9 = 99 and x10 = 100, and the detection rate r (xm) (m = 1,..., 10) for each of them is shown in FIG.
Suppose that it was as shown in FIG. Further, PSNR (Peak Signal to N) is used as the image quality evaluation value.
, Q1 = 20, Q2 = 21,..., Q
31 = 50 (dB). The density function p (x) of the attack parameter selection probability is assumed to be uniform.

Here, the image quality evaluation value frequency distribution value p for each PSNR value Qn and each JPEG image quality parameter value xj
Consider a case where (Qn | xj) is image quality evaluation value frequency distribution information. In this case, the frequency distribution value p (Qn | xj) can be calculated by examining the distribution of PSNR values of the attacked image with respect to each JPEG image quality parameter value xj. FIG. 1 shows an example of actually calculating the distribution using 1000 images.
It is shown in FIG.

For the first detection rate shown in FIG. 12, the image quality evaluation value frequency distribution value p (Q
The second detection rate R (Qn) calculated for each image quality evaluation value Qn using (n | xj) is shown in FIG. Thus, the second detection rate can be calculated.

In the case of the digital watermark characteristic parameter table generation system 70 shown in FIG. 10, the obtained second detection rate is output to the digital watermark characteristic extraction unit 71, where the detection rate description parameter is obtained. . The relationship between the second detection rate and the image quality evaluation value shown in FIG.

[0111]

(Equation 10) By approximation with a logistic curve represented by the following equation, a curve indicated by a broken line in FIG. 15 is obtained.

In this case, the parameters a and b become detection rate description parameters, and in the curve in FIG.
0.347, b = 11.64. Thus, the detection rate description parameter can be calculated.

Next, an embodiment of the detection rate conversion unit 221 shown in FIG. 7 will be described. Here, as in the above embodiment, JPEG compression is considered as an attack. Attack parameter for calculating the first detection rate, image quality evaluation value for calculating the second detection rate, density function p of attack parameter selection probability
The conditions for (x) are the same as in the above embodiment. It is also assumed that the image quality evaluation value attack parameter conversion information is the attack parameter value x'n corresponding to each image quality evaluation value Qn.

The image quality evaluation value attack parameter conversion information stored in the detection rate conversion unit 221 can be calculated, for example, as follows. First, the average value of the distribution shown in FIG. 13 is obtained for each JPEG image quality parameter xj. When actually calculated, the result is as shown in FIG. FIG. 17 shows the relationship between the PSNR values Qn and the corresponding JPEG image quality parameter x′n. This correspondence is stored as image quality evaluation value attack parameter conversion information.

The second detection rate calculated for each image quality evaluation value Qn using the image quality evaluation value attack parameter conversion information shown in FIG. 17 for the first detection rate shown in FIG. R (Qn) is shown in FIG. Thus, the second detection rate can be calculated.

In the case of the digital watermark characteristic parameter table generation system 70 shown in FIG. 10, the obtained second detection rate is output to the digital watermark characteristic extraction unit 71, where the detection rate description parameter is obtained. . When the relationship between the second detection rate and the image quality evaluation value obtained in FIG. 18 is approximated by, for example, a logistic curve represented by Expression 10, a curve indicated by a broken line in FIG. 19 is obtained. In this case, the parameters a and b
Is the detection rate description parameter, and in the curve of FIG.
a = -1.91 and b = 59.56. Thus, the detection rate description parameter can be calculated.

[0117]

As described above, according to the present invention, instead of performing statistical processing of a detection rate for an attack parameter depending on an attack, it is converted into a detection rate for a common index called an image quality evaluation value. By performing the statistical processing to calculate the resistance evaluation value, there is an effect that the resistance evaluation value can be calculated by the common statistical processing regardless of the type of attack. As a result, it is not necessary to design parameters related to statistical processing such as a weighting function for each attack, thereby facilitating the processing. Further, the amount of data to be stored in the tolerance evaluation value calculation unit can be reduced. Further, when the user customizes the resistance evaluation value calculation process, it is not necessary to set the resistance evaluation value calculation parameter for each attack, and the burden on the user can be reduced.

Further, according to the present invention, the degree of attack is converted into an image quality evaluation value which is considered to be directly related to human value judgment, and then the resistance evaluation value is calculated. This has the effect of directly reflecting the value judgment of the user. Since parameters such as a weighting function are directly set for the image quality evaluation value, the obtained tolerance evaluation value is more suitable for human subjective judgment than the conventional one.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of a digital watermark insertion system according to the present invention.

FIG. 2 is a flowchart for explaining the operation of a detection rate conversion unit 13 in FIG.

FIG. 3 is a flowchart illustrating an operation of a tolerance evaluation value calculation unit in FIG. 1;

FIG. 4 is a block diagram illustrating an embodiment of a detection rate conversion unit 13 in FIG.

FIG. 5 is a flowchart for explaining the operation of a detection rate conversion value calculation unit 202 in FIG. 4;

6 is an attack parameter frequency distribution calculation unit 2 in FIG.
13 is a flowchart for explaining the operation of FIG.

FIG. 7 is a block diagram illustrating an embodiment of a detection rate conversion unit 13 in FIG.

8 is a flowchart for explaining the operation of a detection rate conversion unit calculation unit 222 in FIG. 7;

FIG. 9 is a diagram illustrating an example of a digital watermark characteristic table.

FIG. 10 is a block diagram showing another embodiment of the digital watermark insertion system according to the present invention.

FIG. 11 is a flowchart illustrating an operation of a digital watermark characteristic extraction unit.

12 is a graph showing an example of a first detection rate input to the detection rate conversion unit 13 in FIG. 1 or FIG.

13 is a graph illustrating an example of image quality evaluation value frequency distribution information used in an attack parameter frequency distribution calculation unit 203 in FIG.

FIG. 14 is a graph showing an example of a second detection rate output from the detection rate conversion value calculation unit 202 in FIG.

FIG. 15 is a graph showing an example in which a function R (Q) representing a change in the detection rate with respect to the image quality evaluation value is obtained based on the detection rate description parameter calculated from the second detection rate shown in FIG. .

FIG. 16 is a graph showing an example of a relationship between an attack parameter and an image quality evaluation value.

17 is a graph showing an example of image quality evaluation value attack parameter conversion information used by the detection rate conversion value calculation unit 222 in FIG.

18 is a graph illustrating an example of a second detection rate output from the detection rate conversion value calculation unit 222 in FIG.

FIG. 19 is a graph showing an example in which a function R (Q) representing a change in the detection rate with respect to the image quality evaluation value is obtained using the detection rate description parameter calculated from the second detection rate shown in FIG. .

FIG. 20 is a block diagram illustrating an embodiment of a conventional digital watermark insertion system.

FIG. 21 is a flowchart showing an operation of an image quality degradation degree calculation unit in FIG. 20;

FIG. 22 is a flowchart showing the operation of the detection rate calculation unit in FIG.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 storage device 2 digital watermark insertion unit 3 category classification unit 4 digital watermark strength calculation unit 5 input device 10 digital watermark characteristic table generation system 11 image quality degradation calculation unit 12 detection rate calculation unit 13 detection rate conversion unit 14 tolerance evaluation value calculation unit 15 Characteristic Table Generating Unit 50 Digital Watermark Characteristic Calculating Unit 70 Digital Watermark Characteristic Parameter Table Generating System 71 Digital Watermark Characteristic Extracting Unit 75 Characteristic Parameter Table Generating Unit 201 Detection Rate Conversion Unit 202 Detection Rate Conversion Value Calculation Unit 203 Attack Parameter Frequency Distribution Calculation Unit 221 Detection rate conversion unit 222 Detection rate conversion value calculation unit 1210 Digital watermark characteristic table generation system 1211 Resistance evaluation value calculation unit 1212 Characteristic table generation unit

Claims (8)

[Claims] 1. A system for automatically adjusting a digital watermark strength according to an image and inserting a digital watermark, wherein a category amount is calculated by calculating a feature amount of an input image, and a category index data is output. Classification means, an attack parameter which is a parameter for adjusting the degree of attack that can be caused by image processing or image compression on the obtained image by inserting a digital watermark by changing the digital watermark strength for a plurality of images. Is changed, an attack is executed, digital watermark detection is performed on the obtained image, the obtained detection results are totaled, a predetermined category index value, a predetermined digital watermark strength value A detection rate calculating means for calculating a detection rate of a digital watermark for each predetermined attack parameter value and outputting it as first detection rate data; In the first detection rate data, from a series of detection rate data having the same value of the category index and the digital watermark strength, each of the predetermined attacked image quality evaluation values indicating the image quality of the image after the attack is obtained. Using the information describing the relationship with the attack parameters, performing a detection rate conversion to calculate the detection rate for each of the predetermined attacked image quality evaluation value, the predetermined category index value, the predetermined Detection rate conversion means for calculating detection rate data for each of the determined digital watermark strength values, and outputting the calculated detection rate data as second detection rate data; From a series of detection rate data having the same intensity, a detection rate curve representing a change in the detection rate with respect to the attacked image quality evaluation value is obtained, and the detection rate curve is calculated. By performing statistical processing on the digital watermark, a resistance evaluation value is calculated to obtain a resistance evaluation value that is a value representing the resistance of the digital watermark to an attack, and the predetermined category index value and the predetermined digital watermark strength are calculated. A resistance evaluation value calculating unit that calculates a resistance evaluation value for each of the values and outputs the same as second resistance evaluation value data, and inserts a digital watermark by changing the digital watermark strength for a plurality of images. Comparing the obtained image with the image before watermark insertion, the image quality deterioration degree indicating the degree of image quality deterioration of the digital watermark inserted image is determined by the predetermined category index value and the predetermined digital watermark strength value. , And image quality deterioration degree calculating means for outputting as second image quality deterioration degree data, the second resistance evaluation value data and the second image quality deterioration degree data The predetermined category index value, summarized for each of the predetermined digital watermark strength value, is a table describing the relationship between the image quality degradation degree, the robustness evaluation value, the category index and the digital watermark strength, the category index value and A characteristic table generating means for generating and outputting a digital watermark characteristic table, which is a table for obtaining a corresponding resistance evaluation value and a degree of image quality degradation, from the digital watermark strength value, and storing the digital watermark characteristic table, A storage unit for obtaining a corresponding image quality deterioration degree and a robustness evaluation value from the data and the input digital watermark strength data, and outputting the first image quality deterioration degree data and the first robustness evaluation value data; And outputs the obtained digital watermark strength value as the digital watermark strength data to the storage means, and as a result, is output from the storage means. From the first image quality deterioration degree data and the first robustness evaluation value data, an objective function configured using the image quality deterioration degree and the robustness evaluation value based on the input digital watermark strength constraint information is defined. A digital watermark strength calculating means for obtaining a digital watermark strength to be maximized or minimized and outputting the digital watermark strength value as an optimum digital watermark strength value;
A digital watermark insertion unit that inserts the digital watermark signal into the input image using the optimal digital watermark strength value as a parameter, generates and outputs a digital watermark inserted image. 2. A system for automatically adjusting the digital watermark strength according to an image and inserting a digital watermark, wherein the category is calculated by calculating a feature amount of an input image and classifying the input image into category, and outputting category index data. Classification means, an attack parameter which is a parameter for adjusting the degree of attack that can be caused by image processing or image compression on the obtained image by inserting a digital watermark by changing the digital watermark strength for a plurality of images. Is changed, an attack is executed, digital watermark detection is performed on the obtained image, the obtained detection results are totaled, a predetermined category index value, a predetermined digital watermark strength value A detection rate calculating unit that calculates a detection rate for each predetermined attack parameter value and outputs the detection rate data as first detection rate data; In the rate data, the category index, from a series of detection rate data in which the digital watermark strength is the same value, each of the predetermined attacked image quality evaluation values representing the image quality of the image after the attack, and the attack parameter Using the information describing the relationship, the detection rate conversion for calculating the detection rate for each of the predetermined attacked image quality evaluation values is performed, and the predetermined category index value and the predetermined electronic watermark are calculated. A detection rate conversion means for calculating detection rate data for each of the intensity values and outputting the data as second detection rate data; and inserting a digital watermark into the plurality of images by changing the digital watermark strength. The watermarked image is compared with the image before the watermark insertion, and the image quality deterioration degree indicating the degree of image quality deterioration of the digital watermark inserted image is determined by the predetermined category index. Image quality deterioration degree calculating means for calculating each of the index value and the predetermined digital watermark strength value, and outputting the same as the second image quality deterioration degree data. From the same series of detection rate data, a detection rate description parameter calculation for obtaining a detection rate description parameter that is a parameter describing a detection rate surface representing a change in the detection rate with respect to the attacked image quality evaluation value and the digital watermark strength, A description of an image quality degradation degree which is a parameter for describing an image quality degradation degree curve representing a change in the image quality degradation degree with respect to the digital watermark strength from a series of image quality degradation degree data having the same category index in the second image quality degradation degree data Calculating the image quality deterioration degree description parameter for obtaining the parameter, and for each of the predetermined category index values, Calculate the detection rate description parameter and the image quality degradation degree description parameter,
Digital watermark characteristic extracting means for outputting as digital watermark characteristic parameter data; collecting the digital watermark characteristic parameter data for each of the predetermined category index values; and a relationship between the category index, the image quality degradation degree description parameter, and the detection rate description parameter. A characteristic parameter table generating means for creating and outputting a digital watermark characteristic parameter table, which is a table for obtaining an image quality degradation degree description parameter and a detection rate description parameter from a category index value; and A parameter table is accumulated, a corresponding image quality deterioration degree description parameter and a detection rate description parameter are obtained for the category index data, and an image quality deterioration degree curve is obtained from the image quality deterioration degree description parameter for the category index data. Electricity The first image quality deterioration degree data, which is the image quality deterioration degree for the watermark strength data, is calculated, and the relationship between the attacked image quality evaluation value for the digital watermark strength data and the detection rate is determined from the detection rate description parameter for the category index data. , A statistical process is performed on the detection rate curve based on the input resistance evaluation value calculation parameter, and a first resistance evaluation value that is a resistance evaluation value for the digital watermark strength data is obtained. Digital watermark characteristic calculating means for calculating data and outputting the first image quality deterioration degree data and the first resistance evaluation value calculation data; and using the predetermined digital watermark strength value as the digital watermark strength data. Output to the digital watermark characteristic calculating means, and as a result, output from the digital watermark characteristic calculating means. The objective function configured using the image quality deterioration degree and the robustness evaluation value is maximized from the image quality deterioration degree data and the first robustness evaluation value data based on input digital watermark strength constraint information. Or a digital watermark strength calculating means for obtaining a digital watermark strength to be minimized and outputting it as an optimum digital watermark strength value, and converting input embedded data into a digital watermark signal,
A digital watermark insertion unit that inserts the digital watermark signal into the input image using the optimal digital watermark strength value as a parameter, generates and outputs a digital watermark inserted image. 3. A system for automatically adjusting a digital watermark strength according to an image and inserting a digital watermark, wherein the system generates a digital watermark characteristic table used when calculating the digital watermark strength. The digital watermark is inserted into the image by changing the digital watermark strength, and the obtained image is changed by changing an attack parameter which is a parameter for adjusting a degree of an attack that can be caused by image processing or image compression. Performs an attack that can occur due to processing or image compression, performs digital watermark detection on the obtained image, sums up the obtained detection results, and sets a predetermined category index value and a predetermined digital watermark. A detection rate calculator that calculates a detection rate of a digital watermark for each intensity value and a predetermined attack parameter value and outputs the detection rate data as first detection rate data. A predetermined attack image quality evaluation value representing the image quality of an attacked image from a series of detection rate data in which the category index and the digital watermark strength have the same value in the first detection rate data. And, using information describing the relationship between the attack parameters, the detection rate conversion to calculate the detection rate for each of the predetermined attacked image quality evaluation value, the predetermined category index value Detection rate conversion means for calculating detection rate data for each of the predetermined digital watermark strength values, and outputting the calculated detection rate data as second detection rate data; and a category index in the second detection rate data. Calculating a detection rate curve representing a change in the detection rate with respect to the attacked image quality evaluation value from a series of detection rate data having the same digital watermark strength, By performing statistical processing on the fraction curve performs resistant evaluation value calculation for determining the robustness evaluation value is a value representing the resistance to attack of the digital watermark, category index value said predetermined,
A resistance evaluation value calculating unit that calculates a resistance evaluation value for each of the predetermined digital watermark intensity values and outputs the evaluation value as second resistance evaluation value data, and changes the digital watermark intensity for a plurality of images. The obtained image is compared with the image before the watermark is inserted, and the image quality deterioration degree indicating the degree of image quality deterioration of the electronic watermark inserted image is determined by the predetermined category index value and the predetermined image quality. Image quality deterioration degree calculating means for calculating each of the obtained digital watermark strength values and outputting the same as second image quality deterioration degree data; and calculating the second resistance evaluation value data and the second image quality deterioration degree data. And a characteristic table generating means for generating and outputting a digital watermark characteristic table by combining a predetermined category index value and the predetermined digital watermark strength value. Electronic watermark characteristic table generating system. 4. The detection rate conversion means comprises: an attack parameter selection probability information which is information describing a density function of a probability that an attack parameter is selected; and an attack image quality evaluation value under a condition that the attack parameter is constant. Image quality evaluation value frequency distribution information, which is information describing the frequency distribution of the attack parameter selection probability information and the image quality evaluation value frequency distribution information, from the attack parameter and the attacked image quality evaluation value. Determine the ratio of the combined probability to the occurrence probability of the attacked image quality evaluation value, based on the ratio,
Attack parameter frequency distribution calculating means for calculating and outputting attack parameter frequency distribution information, which is information describing the frequency distribution of attack parameters under the condition that the image quality evaluation value of the attacked image is constant, from the attack parameter frequency distribution information An attack parameter frequency distribution for each of the predetermined attacked image quality evaluation values is determined, and in the first detection rate data, the attack index is calculated from a series of detection rate data having the same category index and digital watermark strength. A detection rate transition function representing a change in the detection rate with respect to the parameter is obtained, and a value based on a product of the attack parameter frequency distribution and the detection rate transition function is integrated with respect to the attack parameter, thereby obtaining the predetermined value. Calculate the detection rate for each attacked image quality evaluation value A detection rate conversion value is calculated by performing detection rate conversion, calculating detection rate data for each of the predetermined category index value and the predetermined digital watermark strength value, and outputting the data as second detection rate data. 4. The digital watermark characteristic table generation system according to claim 3, comprising: a calculation unit. 5. The detection rate conversion means has image quality evaluation value attack parameter conversion information describing an image quality evaluation value attack parameter conversion function, which is a function of converting an attacked image quality evaluation value into an attack parameter, Calculating the image quality evaluation value attack parameter conversion function from the image quality evaluation value attack parameter conversion information, calculating an attack parameter value corresponding to each of the predetermined attacked image quality evaluation values to obtain an equivalent attack parameter value, Among the detection rate data, the category index, from a series of detection rate data having the same digital watermark strength, a detection rate transition function representing a change in the detection rate with respect to the attack parameter is obtained, and the detection rate transition function is used. The detection rate for each of the equivalent attack parameter values is By performing detection rate conversion to be calculated as the detection rate of the attacked image quality evaluation value obtained, the predetermined category index value, the detection rate data is calculated for each of the predetermined digital watermark strength value, 4. The digital watermark characteristic table generation system according to claim 3, further comprising a detection rate conversion value calculating unit that outputs the detection rate conversion value data as second detection rate data. 6. A system for automatically adjusting a digital watermark strength according to an image and inserting a digital watermark, wherein the digital watermark characteristic parameter table used for calculating the digital watermark strength is generated. By changing the digital watermark strength for the image and inserting a digital watermark, for the obtained image, changing the attack parameter, which is a parameter for adjusting the degree of attack that can be caused by image processing or image compression, Performs an attack that can occur by image processing or image compression, detects digital watermarks on the obtained images, aggregates the obtained detection results, and sets a predetermined category index value, a predetermined electronic A digital watermark detection rate is calculated for each of the watermark strength value and a predetermined attack parameter value, and is output as first detection rate data. A detection rate calculating means, a predetermined attack image representing the image quality of the image after the attack, from a series of detection rate data having the same value of the category index and the digital watermark strength in the first detection rate data. Using the information describing the relationship between each of the image quality evaluation values and the attack parameters, the detection rate conversion for calculating the detection rate for each of the predetermined attacked image quality evaluation values is performed. A detection rate conversion unit that calculates detection rate data for each of the category index value and the predetermined digital watermark strength value and outputs the detection rate data as second detection rate data; The digital watermark is inserted by changing it, and the obtained image is compared with the image before the watermark is inserted. Image quality deterioration degree calculating means for calculating each of a predetermined category index value and the predetermined digital watermark strength value and outputting the second image quality deterioration degree data; From a series of detection rate data having the same category index, a detection rate description parameter which is a parameter describing a detection rate surface representing a change in the detection rate with respect to the attacked image quality evaluation value and the digital watermark strength. A parameter describing a rate description parameter calculation and an image quality deterioration degree curve representing a change in image quality deterioration degree with respect to digital watermark strength from a series of image quality deterioration degree data having the same category index in the second image quality deterioration degree data. And the image quality deterioration degree description parameter is calculated to obtain the image quality deterioration degree description parameter. Digital watermark characteristic extraction means for calculating a detection rate description parameter and an image quality degradation degree description parameter for each of the index values and outputting the digital watermark characteristic parameter data as the digital watermark characteristic parameter data; And a characteristic parameter table generating means for generating and outputting a digital watermark characteristic parameter table by summarizing for each index value. 7. The detection rate conversion means, comprising: an attack parameter selection probability information which is information describing a density function of a probability that an attack parameter is selected; and an attack image quality evaluation value under a condition that the attack parameter is constant. Image quality evaluation value frequency distribution information, which is information describing the frequency distribution of the attack parameter selection probability information and the image quality evaluation value frequency distribution information, from the attack parameter and the attacked image quality evaluation value. Determine the ratio of the combined probability to the occurrence probability of the attacked image quality evaluation value, based on the ratio,
Attack parameter frequency distribution calculating means for calculating and outputting attack parameter frequency distribution information, which is information describing the frequency distribution of attack parameters under the condition that the image quality evaluation value of the attacked image is constant, from the attack parameter frequency distribution information An attack parameter frequency distribution for each of the predetermined attacked image quality evaluation values is determined, and in the first detection rate data, the attack index is calculated from a series of detection rate data having the same category index and digital watermark strength. A detection rate transition function representing a change in the detection rate with respect to the parameter is obtained, and a value based on a product of the attack parameter frequency distribution and the detection rate transition function is integrated with respect to the attack parameter, thereby obtaining the predetermined value. Calculate the detection rate for each attacked image quality evaluation value A detection rate conversion value is calculated by performing detection rate conversion, calculating detection rate data for each of the predetermined category index value and the predetermined digital watermark strength value, and outputting the data as second detection rate data. 7. The digital watermark characteristic parameter table generating system according to claim 6, comprising: a calculating unit. 8. The detection rate conversion unit has image quality evaluation value attack parameter conversion information that describes an image quality evaluation value attack parameter conversion function that is a function of converting an attacked image quality evaluation value into an attack parameter. Calculating the image quality evaluation value attack parameter conversion function from the image quality evaluation value attack parameter conversion information, calculating an attack parameter value corresponding to each of the predetermined attacked image quality evaluation values to obtain an equivalent attack parameter value, Among the detection rate data, the category index, from a series of detection rate data having the same digital watermark strength, a detection rate transition function representing a change in the detection rate with respect to the attack parameter is obtained, and the detection rate transition function is used. The detection rate for each of the equivalent attack parameter values is By performing detection rate conversion to be calculated as the detection rate of the attacked image quality evaluation value obtained, the predetermined category index value, the detection rate data is calculated for each of the predetermined digital watermark strength value, 7. The digital watermark characteristic parameter table generation system according to claim 6, further comprising a detection rate conversion value calculating unit that outputs the detection rate data as second detection rate data.