JP2005217457A - Electronic watermark combining device, electronic watermark combining program, authentication information determining device, and authentication information determininig program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To embed an electronic watermark and extract it so that the electronic watermark can be extracted without tracing back procedures used for combining and the combined electronic watermark is hardly removed even if compression processing, filtering or geographic modification is applied. <P>SOLUTION: A device is provided with an independent evaluating section 5, a watermark vector correcting section 6 and a linear coupling vector generating section 7. The device corrects image data in an embedding section 8 using a linear coupling vector obtained by linearly coupling an image vector with a watermark vector which are independent from each other to combine the electronic watermark. With this configuration, in extracting the electronic watermark from the combined image data after a mixed vector is obtained, the watermark vector is separated from the mixed vector by a technique of independent component analysis, and authentication information is determined without tracing back the the procedures used for combining. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a digital watermark embedding apparatus and method for synthesizing a digital watermark with digital content such as an image and audio, a digital watermark extraction apparatus for extracting authentication information by extracting a digital watermark embedded from the synthesized digital content, and In particular, the present invention relates to an image data processing apparatus that performs authentication using independent component analysis for separating a signal from an observation signal in which a plurality of signals are mixed based on statistical independence.

  In recent years, in addition to digitization of images such as broadcasts and movies, digital devices for consumer use have become cheaper and higher performance, and digital images have become familiar to many people. In addition, with the broadbandization of the Internet and the rapid spread of mobile phones with cameras, it is expected that the distribution of digital video content will become full-scale in the future. Under such circumstances, research on digital watermarking technology that embeds data such as copyright information in digital data so that it is not perceived by human beings is actively conducted as a method for protecting the copyright of digital contents such as images and audio. (See Non-Patent Document 1 and Non-Patent Document 2).

Typical methods for synthesizing a digital watermark with an image include, for example, a method of directly embedding authentication information in a pixel value (spatial domain) of an image, and authentication information in a transform coefficient (frequency domain) of discrete cosine transform or wavelet transform. Etc., and a number of synthesis methods have been proposed for either method. When an original digital watermark is extracted from an image synthesized by these methods, in principle, any method is performed by following the procedure used for synthesis in reverse.
Kokoo Matsui, “Basics of Digital Watermarking” Morikita Publishing, August 1998 "Nikkei Electronics" Nikkei BP, February 24, 1997 (P149-162)

  However, if compression processing, filtering, or geometrical modification is added after synthesizing the digital watermark, the original digital watermark cannot be extracted simply by following the procedure used for synthesis in reverse. Furthermore, in the technique of performing extraction by reversing the procedure of synthesizing the digital watermark, by analyzing the process of extracting the digital watermark, it is estimated where and how the digital watermark is synthesized in the image data. There is also a possibility that the digital watermark is removed and the fraudulent use is performed while minimizing.

  The present invention has been made in view of the above-described problems of the prior art, and can extract a digital watermark without reversing the procedure used for synthesis, and can be applied with compression processing, filtering, and geometric modification. An object of the present invention is to provide a digital watermark synthesizing apparatus and method and a digital watermark extracting apparatus and method in which the synthesized digital watermark is difficult to be removed.

  In order to achieve the above object, the synthesizing device according to the present invention generates authentication information (digital watermark data) that is statistically independent of digital content to be authenticated, and then combines the two by linearly combining them. Separation is performed using the property that the synthesized image content and the authentication information are statistically independent.

  Before specifically describing the means for solving the above-mentioned problems, first, an outline of independent component analysis will be described.

  Assume that there are m types of vectors s1, s2,..., sm, and the vector s is given by Equation 1.

s = (s1, s2, ..., sm) ^T (Formula 1)
Here, T represents transposition. Further, it is assumed that each component of s has an average value of 0. At this time, when the components of s are statistically independent from each other, Equation 2 is established.

P (s) = P (s1) P (s2)... P (sm) (Formula 2)
Here, the probability distribution P (s) represents the probability distribution of the vector s, and the probability distribution P (sk) (1 ≦ k ≦ m) represents the probability distribution of the k-th component vector sk of the vector s.

  On the other hand, a vector x composed of n types of vectors x1, x2,..., Xn obtained by linearly combining the component vectors of Expression 1 is defined by Expression 3.

x = (x1, x2,..., xn) ^T (Equation 3)
Then, the relationship of Formula 4 is established.

x = As (Formula 4)
Here, the linear operator A represents the degree of mixing of the component vectors s1, s2,. Independent component analysis means that the mixed vector x is obtained by Equation 5 using only the fact that the component vectors s1, s2,... Is separated into vectors y independent of each other before mixing.

y = Wx (Formula 5)
Here, the vector y is expressed by Equation 6, and the component vectors y1, y2,..., Ym correspond to any of the vectors s1, s2,.

y = (y1, y2, ..., ym) ^T (Formula 6)
Based on this, in order to solve the above problems, the digital watermark synthesis apparatus of the present invention is:
A digital watermark synthesis apparatus that embeds authentication information data in digital data by linearly combining L vectors (L ≧ 2) with each other, and means for acquiring vector v1 from the digital data, and the authentication information When the function F (v1, v2) is greater than or less than a threshold using means for obtaining a vector v2 from data and a function F (v1, v2) for evaluating the independence of the vector v1 and the vector v2, Means for repeatedly correcting the vector v1 or the vector v2 until the function F (v1, v2) is less than or equal to the threshold value, and linearly combining the vector v1 and the vector v2 to newly add a vector v3 and a vector v4. Generating the authentication information data to the digital data using the vector v3 and the vector v4. And means for obtaining a watermark composite digital data embedded as a watermark.

Furthermore, in order to solve the above-described problem, the authentication information determination device of the present invention provides:
From the digital data, m types (m ≧ 2) of vectors w1, w2,. . . , Wm, and the vectors w1, w2,. . . , Wm are subjected to independent component analysis, and n types (2 ≦ n ≦ m) of restoration vectors x1, x2,. . . , Xn, and the restoration vectors x1, x2,. . . , Xn, respectively, and normalization restoration vectors y1, y2,. . . , Yn and the normalized restoration vectors y1, y2,. . . , Yn, respectively, means for generating first to n-th separated data, and means for selecting digital watermark data from the first to n-th separated data.

  With such a processing configuration, digital watermark embedding can be performed as follows. So that statistical independence is ensured by modifying the digital data vector generated from the digital data or the digital watermark data vector generated from the authentication information data so that it is not visually perceived by mixing by linear combination; and It can be embedded in digital data so that digital watermark data can be separated from the mixed vector by independent component analysis.

  In addition, digital watermark data can be extracted as follows. m types (m ≧ 2) of vectors w1, w2,. . . After acquiring wm, as a general blind signal source separation problem, the parameter information for vector separation is not required at all by the general independent component analysis method without following the reverse of the digital watermark embedding procedure. It is possible to extract a digital watermark without using it.

  According to the electronic watermark synthesizing apparatus or the computer program for causing the computer to execute the electronic watermark synthesizing means, statistical independence is evaluated, and the digital data vector generated from the digital data or the electronic data generated from the authentication information data By modifying the watermark data vector, the statistically independent digital data vector and the watermark vector are mixed by a linear combination so as not to be visually perceived, and from the mixed vector, an electronic component is analyzed by independent component analysis. It can be embedded in digital data so that the watermark data can be separated.

  Further, according to the authentication information determination apparatus or the program for causing a computer to execute the authentication information determination means of the present invention, m types (m ≧ 2) of vectors w1, w2,. . . After acquiring wm, as a general blind signal source separation problem, the parameter information for vector separation is not required at all by the general independent component analysis method without following the reverse of the digital watermark embedding procedure. Without the need to extract the digital watermark and prevent the parameter information used for embedding and extracting the digital watermark data from being leaked, estimating how the digital watermark is synthesized, removing the digital watermark, and The possibility of being used can be reduced.

  In addition, even after compression processing, filtering, geometric modification, etc. are added to digital data after synthesizing the digital watermark, the change vector due to the modification can be separated by independent component analysis. Can increase the sex.

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

(Embodiment)
Here, digital data for embedding a digital watermark is image data of horizontal M pixels and vertical N pixels, and an image of one-sixth of the image to be embedded with the digital watermark is synthesized as authentication information data. For example, the resolution of the image data used as the authentication information may be an arbitrary size. Further, the digital data to be embedded with the digital watermark does not have to be an image, and may be audio, for example. The authentication information data is not limited to image data.
FIG. 1 shows a processing block diagram of the digital watermark embedding method of the present invention. In FIG. 1, the image data input unit 1 captures image data for embedding a digital watermark, and the image vector generation unit 2 acquires an image vector from the image data captured by the image data input unit 1.
The watermark image data input unit 3 captures watermark image data used as authentication information data, and the watermark vector generation unit 4 acquires a watermark vector from the watermark image data captured by the watermark image data input unit 3.
The independence evaluation unit 5 evaluates the independence of the image vector and the watermark vector, and the watermark vector correction unit 6 corrects the watermark vector when the evaluation value in the independence evaluation unit 5 is equal to or greater than a threshold value. When the independence evaluation value is less than the threshold value, the linear combination vector generation unit 7 acquires two types of new linear combination vectors by linearly combining the image vector and the watermark vector.
The embedding unit 8 uses the generated two types of linear combination vectors to modify the image data input from the image data input unit 1 to embed a digital watermark, and the synthesized image data output unit 9 converts the digital watermark image. Output the combined image.

  About the process block diagram comprised as mentioned above, the operation | movement is demonstrated using the flowchart shown in FIG.

  First, the image data input unit 1 and the watermark image data input unit 3 read out M × N pixel image data for embedding a digital watermark and M ′ × N ′ pixel watermark image data for embedding as authentication information, respectively (S1). ). Here, M ′ = M / 6 and N ′ = N / 6 (the operator / represents the integer part of the quotient).

  Next, an image vector v1 is acquired using a conversion coefficient obtained by wavelet transform of the read image data (S2 to S3). The specific procedure is as follows.

  First, wavelet transform is performed on image data in units of 2 × 2 pixels. The wavelet transform for the image is realized by scanning the image signal with a one-dimensional low-pass filter and a one-dimensional high-pass filter in the horizontal direction and the vertical direction, respectively. As a result, the image is divided into four multi-resolution components (FIG. 3). In FIG. 3, LL corresponds to a low frequency component, LH corresponds to a high frequency component in the vertical direction, HL corresponds to a high frequency component in the horizontal direction, and HH corresponds to a high frequency component in the oblique direction. The inverse transformation is realized by scanning the one-dimensional low-pass filter and the one-dimensional high-pass filter in the order of the vertical direction and the horizontal direction with respect to the transform coefficient after the wavelet transform. Thereby, a restored image having the same resolution as the original image can be obtained.

  Next, among the obtained transform coefficients, an LL component image using only the LL component is acquired (S2), and the LL component image is divided into blocks of 3 × 3 pixels, and the horizontal block from the upper left block The center pixel (coordinates (3m + 1, 3n + 1)) of the m-th (0 ≦ m <M / 2) and n-th (0 ≦ n <N / 2) vertical 3 × 3 pixel block is the (m + n × M) th. ') An image vector v1 as a component is acquired (S3).

  Next, the brightness value of the watermark image data input to the watermark image data input unit 3 is acquired, and the watermark vector having the brightness value at coordinates (m ′, n ′) as the (m ′ + n ′ × M ′) component v2 is generated (S4).

  Next, the independence of the image vector v1 and the watermark vector v2 obtained in steps S2 to S4 is evaluated (S5). The evaluation function is defined as shown below based on Equation 2.

F (v1, v2) = ΣΣP _ij (v1, v2) log ₂ P _ij (v1, v2)
-ΣP _i (v1) log ₂ P _i (v1)
-ΣP _i (v2) log ₂ P _i (v2) (Formula 7)
Here, the probability P _ij (v1, v2) represents the probability that the combination of the component values of the image vector v1 and the watermark vector v2 is i and j. Probability Pi (v1) and probability Pi (v2) represent the probability that the component value of image vector v1 and the component value of watermark vector v2 are i, respectively.

  Next, the obtained evaluation value is compared with a threshold value (S6), and when the evaluation value is equal to or larger than the threshold value, the watermark vector v2 is corrected (S7).

  The processing from step S5 to step S7 is repeated until the evaluation value becomes less than the threshold value. Specifically, it is as follows.

  First, correction of the vector v2 in the K-th process is performed only for the k-th component (0 ≦ k <M ′ × N ′) determined by Expression 8 among the components of the vector v2, and the correction value v2k ′ is expressed by Expression 9. Define.

k = (K / (2 × v + 1))% (M ′ × N ′) (Formula 8)
v2 _{k ′} = v2 _kORG + ((K−1)% (2 × v + 1)) − v (formula 9)
Here, v2 _kORG is the k-th component of the watermark vector v2 acquired in step S4, v is the upper limit value of the correction width of the elements of the vector v2, and the operator% represents an operator for calculating the remainder.

Next, an evaluation value when the modified value v2 _{k ′} vector v2 is the k-th component is calculated. If the evaluation value is equal to or greater than the calculated minimum evaluation value, the k-th component of the vector v2 is a value before calculating the evaluation value. If the calculated value is smaller than the calculated minimum evaluation value, the k-th component of the vector v2 is replaced with the modified value v2 _{k ′} and the calculated minimum evaluation value is replaced with the calculated evaluation value.

In these processes, when the calculated minimum evaluation value becomes less than the threshold value, the process from step 5 to step 7 is terminated, and two types of vectors v3 obtained by linearly combining the image vector v1 and the modified watermark vector v2; A vector v4 is generated (S8). The k-th components of the vectors v3 and v4 at that time are set as follows using the k-th components v1 _k and v2 _k of the vectors v1 and v2.

v3 _k = (1 + α) × v1 _k −β × v2 _k (Formula 10)
v4 _k = (1−α) × v1 _k + β × v2 _k (Formula 11)
Here, α is an embedding strength, and β is a parameter for suppressing deterioration in image quality.

  Finally, the image data is corrected using the vector v3 and the vector v4. The procedure is as follows.

First, the value of the coordinates (3m + 1, 3n + 1) of the LL component image is replaced with the value of the (m + n × M ′) component of the vector v3 (S9). Next, the difference between the value of the (m + n × M ′) component of the vector v4 and the average value of the neighboring 8 pixels of the coordinates (3m + 1, 3n + 1) of the LL component image is added to the neighboring 8 pixels (S10). Next, the corrected LL image data is replaced with the LL component after wavelet transformation, and inverse wavelet transformation is performed every 2 × 2 pixels (S11). Finally, the obtained composite image data is output (S12).
Next, a processing example of a program for realizing the authentication information determination device or the authentication information determination means will be described.

FIG. 4 shows a processing block diagram of a program for executing the authentication information determination apparatus or the authentication information determination means of the present invention. The composite image data input unit 11 takes in composite image data in which a digital watermark is combined, and a mixed vector. The acquisition unit 12 acquires two types of mixed vectors from the composite image data captured by the composite image data input unit 11.
The vector separation unit 13 performs independent component analysis on the two types of mixed vectors acquired by the mixed vector acquisition unit 12 and separates and acquires two types of mutually independent vectors. The vector normalization unit 14 normalizes the two types of mutually independent vectors acquired by the vector separation unit 13 in order to solve the arbitraryness of the size of the vector element obtained by the independent component analysis, and reconstructs the image data The unit 15 reconstructs image data from two types of normalized vectors that are independent of each other.
The watermark image selection unit 16 selects a watermark image from the two types of reconstructed image data by comparing with the composite image, and the authentication information determination unit 17 determines whether the selected watermark image data is authentication information. Is determined using the data input from the authentication information determination data input unit 18.

About the process block diagram comprised as mentioned above, the operation | movement is demonstrated using the flowchart shown in FIG.
First, composite image data of M × N pixels combined with a watermark image is read out to the composite image data input unit 11 (S101). Next, the luminance component of the composite image data is wavelet transformed in units of 2 × 2 pixels, and the LL component is acquired from the obtained conversion coefficients to construct the LL component image (S102). After this LL component image is divided into blocks of 3 × 3 pixels, an image vector w1 having the center pixel (coordinates (3m + 1, 3n + 1)) of the 3 × 3 pixel block as the (m + n × M ′) component is obtained. (S103).
Next, an image vector w2 having an average value of 8 pixels in the vicinity of the coordinates (3m + 1, 3n + 1) as the (m + n × M ′) component is obtained (S104), and the image vector w1 and the image vector w2 are subjected to independent component analysis. Two types of mutually independent vectors x1 and x2 are acquired (S105). Any algorithm of independent component analysis may be used (for example, see Shunichi Amari and Noboru Murata, “Independent component analysis: a new method of multivariate data analysis”, Science (November 2002)).
Since the vector x1 and the vector x2 obtained by the independent component analysis have arbitrary element sizes, in order to solve this problem, normalize x1 and x2 based on the luminance distribution of the LL component image, respectively. Thus, vector y1 and vector y2 are obtained (S106). Specifically, the normalization method is as shown in Equation 12.

ya = (254 * (xa-MINa)) / (MAXa-MINa) +1 (Formula 12)
Here, a = 1 and 2, and MINa and MAXa respectively represent the minimum value and the maximum value among the elements of the vector xa.
Then, image data I1 and image data I2 of two types of M ′ × N ′ pixels having the luminance value of the coordinates (m, n) as the (m + n × M ′) component of the obtained vector y1 and vector y2, respectively. Reconstruction is performed (S107).
Next, image data I1, image data I2, and LL component image data are displayed and compared, and image data different from the LL component image is selected as watermark image data (S108).

Finally, the selected watermark image data and authentication information image data are displayed side by side to determine whether or not the selected watermark image data is authentication information (S109).
As described above, according to the embodiment of the present invention, the vector obtained from the image data and the watermark image is corrected so as to be statistically independent from each other, and then linearly combined and combined to be visually perceived. In addition, the watermark image data can be embedded in the image data so as to be separated from the synthesized image data by independent component analysis.
Furthermore, after obtaining the two types of vectors w1 and w2 from the composite image data, the extraction of the watermark image can be attributed to the general blind signal source separation problem, and without following the reverse of the digital watermark embedding procedure, With a general independent component analysis technique, it is possible to extract a digital watermark without requiring any parameter information for separating two types of vectors.
Further, even when compression processing, filtering, or geometric modification is applied after synthesizing a digital watermark, a change vector of pixel values due to the modification can be separated from an image vector or a watermark image vector by independent component analysis.
As a result of the experiment, as a result of embedding watermark image data with α = 0.025 and β = 0.05 with respect to a grayscale image of 960 × 960 pixels (M = N = 960) and 256 gradations, the image quality is visually determined. I couldn't perceive the deterioration of. In addition, it was possible to extract a watermark vector from image data obtained by adding noise, JPEG compression, or partial cropping to the synthesized image data obtained by synthesizing the watermark image data.

  In the present embodiment, an example in which a digital watermark is synthesized with LL component coefficients after wavelet transform of an image has been described. However, synthesis may not be performed with an LL component, and the block size on which wavelet transform is performed. The block size of the process for synthesizing the digital watermark may be an arbitrary size. In addition to the wavelet transform conversion coefficient, the electronic watermark may be combined directly with the pixel value of the image data, or may be combined with a conversion coefficient other than the wavelet transform such as discrete cosine transform. Further, the independence evaluation function, the vector acquisition method used for synthesis, and the vector synthesis method may be methods other than those shown in the present embodiment, and watermark image data selection and authentication information determination are performed by displaying images. In addition to the method of selecting and determining by a human, a method of automatically determining everything without human intervention or a method of automatically determining one and determining the other automatically may be used. .

  The digital watermark synthesis apparatus for embedding authentication information data in digital data by linearly combining mutually independent vectors according to the present invention, or the authentication information data being embedded in digital data by linearly combining mutually independent vectors in a computer A program for executing the means and an authentication information determination device based on digital watermark extraction using independent component analysis, or a program for causing a computer to execute the authentication information determination means evaluates the statistical independence and performs digital data The digital data vector generated from the digital data or the watermark data vector generated from the authentication information data is modified so that the digital data vector and the watermark vector which are statistically independent are mixed by linear combination and visually perceived. Not mixed and mixed Can be embedded in digital data so that digital watermark data can be separated from the vector by independent component analysis, and without following the reverse of the digital watermark embedding procedure, without any parameter information for vector separation, Digital watermark extraction can be performed by a general independent component analysis method, and parameter information used for embedding and extracting digital watermark data can be prevented from leaking, and how the digital watermark is synthesized can be estimated. The watermark can be removed and the possibility of unauthorized use can be reduced. Furthermore, even if digital data is subjected to compression processing, filtering, geometric modification, etc. By component analysis, change vectors due to modification can be separated and the reliability of the separated digital watermark data is improved. The digital watermark embedding device that synthesizes a digital watermark with digital content such as images and audio, or a program that causes a computer to execute a digital watermark embedding means, and the digital watermark embedded from the synthesized digital content are extracted and authenticated It is useful as a digital watermark extraction apparatus for determining information or a program for causing a computer to execute digital watermark extraction means.

Processing configuration diagram of digital watermark embedding method of the present invention The flowchart which shows the process of the digital watermark embedding method of this invention Illustration of wavelet transform coefficient Processing configuration diagram of digital watermark extraction method of the present invention The flowchart which shows the process of the digital watermark extraction method of this invention

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Image data input part 2 Image vector generation part 3 Watermark image data input part 4 Watermark vector generation part 5 Independence evaluation part 6 Watermark vector correction part 7 Linear combination vector generation part 8 Embedding part 9 Composite image data output part 11 Composite image data Input unit 12 Mixed vector acquisition unit 13 Vector separation unit 14 Vector normalization unit 15 Image data reconstruction unit 16 Watermark image selection unit 17 Authentication information determination unit 18 Authentication information determination data input unit

Claims (8)

A digital watermark synthesis apparatus for embedding authentication information data in digital data by linearly combining L sets (L ≧ 2) of mutually independent vectors, the means for acquiring a vector v1 from the digital data, and the authentication information The function F (v1, v2) is greater than or less than a threshold value by using means for obtaining a vector v2 from data, the vector v1, and the function F (v1, v2) for evaluating the independence of the vector v2. In this case, until the function F (v1, v2) is equal to or lower than the threshold value, the vector v1 or the means for correcting the vector v2, and the vector v1 and the vector v2 are linearly combined. The vector v3, the means for generating the vector v4, the vector v3, and the vector v4 are used for the digital data. The witness information data electronic watermark synthesis and means for obtaining a watermark composite digital data embedded as an electronic watermark. An electronic watermark synthesizing program for causing a computer to execute means for embedding authentication information data in digital data by linearly combining L sets (L ≧ 2) of mutually independent vectors, the vector v1 from the digital data Using the function F (v1, v2) for evaluating the independence of the vector v1, the vector v2, and the vector v2 from the authentication information data, the function F (v1, v2) if v2) is greater than or equal to a threshold or less, means for repeatedly modifying the vector v1 or the vector v2 until the function F (v1, v2) is less than or greater than or equal to the threshold; Means for linearly combining the vector v1 and the vector v2 to generate a new vector v3 and vector v4; the vector v3; Using vector v4, watermarking synthesis program for the verification data to execute means for acquiring electronic watermark composite digital data embedded as an electronic watermark in the digital data. From the digital data, m types (m ≧ 2) of vectors w1, w2,. . . , Wm, and the vectors w1, w2,. . . , Wm are subjected to independent component analysis, and n types (2 ≦ n ≦ m) of restoration vectors x1, x2,. . . , Xn, and the restoration vectors x1, x2,. . . , Xn, respectively, and normalization restoration vectors y1, y2,. . . , Yn and the normalized restoration vectors y1, y2,. . . , Yn, respectively, means for generating first to nth separated data, and means for selecting digital watermark data from the first to nth separated data. The authentication according to claim 3, wherein the digital data is synthesized digital data synthesized by the digital watermark synthesizing apparatus according to claim 1 or the digital watermark synthesizing program according to claim 2. Information determination device. 5. The authentication information determining apparatus according to claim 3, further comprising means for determining whether or not the selected digital watermark data is authentication information data. The computer receives m types (m ≧ 2) of vectors w1, w2,. . . , Wm, and the vectors w1, w2,. . . , Wm are subjected to independent component analysis, and n types (2 ≦ n ≦ m) of restoration vectors x1, x2,. . . , Xn, and the restoration vectors x1, x2,. . . , Xn, respectively, and normalization restoration vectors y1, y2,. . . , Yn and the normalized restoration vectors y1, y2,. . . , Yn, an authentication information determination program for executing means for generating first to n-th separated data and means for selecting digital watermark data from the first to n-th separated data, respectively. The authentication according to claim 6, wherein the digital data is synthesized digital data synthesized by the digital watermark synthesizing apparatus according to claim 1 or the digital watermark synthesizing program according to claim 2. Information judgment program. 8. An authentication information determination program according to claim 6, further comprising means for determining whether or not the selected digital watermark data is authentication information data. program.

Images