JP2009141554A - Method and device for embedding electronic watermark and device for detecting electronic watermark - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for embedding electronic watermarks which have durability, as well as, have block-to-block boundaries that are made inconspicuous. <P>SOLUTION: An image is divided into blocks and confidential information is embedded in a block (A<SB>ij</SB>). A shift value α is set for the block (A<SB>ij</SB>) to change the average value of pixels in the block according to the confidential information embedded in the block. Each pixel value in the block is increased or decreased by a uniform random number, according to the shift value α so that the average pixel value (<A1<SB>ij</SB>>) of the block, after the confidential information is embedded and the average pixel value (<A0<SB>ij</SB>>) of the block before the confidential information is embedded, satisfy the relation in formula (1): formula (1) (<A1<SB>ij</SB>>)≥ä(<A0<SB>ij</SB>>)+α}where (<A0<SB>ij</SB>>) and (<A1<SB>ij</SB>>) are the respectively average pixel values of the blocks A0<SB>ij</SB>and A1<SB>ij</SB>before and after information embedding and the shift value α is positive, when the confidential information is 1 or negative, when the confidential information is 0. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a method of embedding secret information as an electronic watermark in an image, an electronic watermark embedding device using the method, and an electronic watermark detection device.

  Since digitized still images and moving images can be easily duplicated without image deterioration, copyright is a problem. Therefore, as one of the measures for protecting the copyright, a digital watermark technique for embedding copyright information as confidential information in an image is used.

  In digital watermarks applied to copyright protection, the embedded secret information cannot be perceived simply by looking at the image, the deterioration of the image due to the embedded confidential information cannot be perceived, and various processing is applied to the image. However, it is required that the embedded secret information is not easily destroyed, and an embedding technique capable of satisfying these requirements is demanded.

  One method of embedding confidential information such as copyright information in an image is to divide the image into small blocks and embed 1-bit confidential information (“1” or “0”) in blocks. Yes. As a method of embedding secret information in that case, a method of expressing 1-bit secret information using an average value of pixel values (for example, luminance values) of each block is known (for example, see Non-Patent Document 1). .

  FIG. 12 is a diagram for explaining the principle of a method for embedding secret information and a method for detecting secret information using an average of pixel values for each block. An original image before embedding secret information is set as a reference image, and an image after embedding secret information is set as an embedded image. Here, description will be made assuming that the pixel value is luminance information (Y) of the (Y, Cb, Cs) color system, but if it is pixel information, the three primary color information expressed by (R, G, B), (Y , Cb, Cs) luminance information and color information based on the color system, and (Y, I, Q) luminance information and color information based on the color system.

For convenience of explanation, it is assumed that the entire image is composed of nine blocks arranged three by three vertically and horizontally. i is a block number in the row direction, j is a block number in the column direction, a block of the reference image is called R _ij, and an embedded image is called S _ij . Therefore, in FIG. 11, the blocks R ₁₁ , R ₁₂ , R ₃₁ , and R ₃₂ on the reference image side have a one-to-one correspondence with the blocks S ₁₁ , S ₁₂ , S ₃₁ , and S 32 on the embedded image side where the secret information is embedded, respectively. Corresponding to

Further, for convenience of explanation, each block is composed of a pixel group of 4 rows × 4 columns, the blocks R ₁₁ and R ₁₂ have a large dispersion of pixel values (a large variation in pixel values), and the blocks R ₃₁ and R ₃₂ have It is assumed that the variance of pixel values is the smallest (the pixel values are all constant).

As a rule for embedding 1-bit confidential information (“1” or “0”) for each block, for the block in which “1” is embedded, the average pixel value of the block is shifted (changed) to a value larger than the original image. For the block in which “0” is embedded, the average of the pixel values of the block is shifted (changed) to a value smaller than the original image. That is, assuming that the average of the pixel values of the blocks S _ij and R _{ij in} the i row and the j column is <S _ij > and <R _ij >, the average of the pixel values is increased or decreased by the shift value α (≠ 0). An operation is performed on each pixel in the block so that the equation is established.
<S _ij > = <R _ij > + α
However, α when embedding “1” is a positive integer, and α when embedding “0” is a negative integer.

At this time, the average is shifted by α by uniformly adding or subtracting the shift value α to each pixel value constituting the block A _ij to be embedded.
For example, the shift value alpha = 2, as shown in FIG. 12 (a), the block _{R 11,} against _{R 31,} to embed secret information "1", uniformly adds the 2 _{S 11,} and S _31. In order to embed the secret information “0” for the blocks R ₁₂ and R ₃₂ , 2 is uniformly subtracted to obtain S ₁₂ and S ₃₂ .

In this case, the average pixel values <R ₁₁ > and <R ₃₁ > of the blocks R ₁₁ and R ₃₁ are 4.8125 and 5, respectively (<R ₁₁ > is (5 + 3 + 10 +... + 8 + 5 + 7) /16=4.8125). The average pixel values <S ₁₁ > and <S ₃₁ > of the blocks S ₁₁ and S ₃₁ are 6.8125 and 7, respectively, which are two larger than <R ₁₁ > and <R ₃₁ >. Similarly, the average pixel values <R ₁₂ > and <R ₃₂ > of the blocks R ₁₂ and R ₃₂ are 5.625 and 5, respectively, and the average pixel values of the blocks S ₁₂ and S ₃₂ are <S ₁₂ > and <S _32. > Is 3.625, 3 which is smaller by 2 than <R ₁₂ > and <R ₃₂ >, respectively.

On the other hand, when detecting the secret information from the embedded image in which the secret information is embedded, the image is divided into the same blocks as at the time of embedding and the blocks R ₁₁ , R ₁₂ , R ₃₁ , R ₃₂ , S ₁₁ , S ₁₂ , S _{31 are used.} for each of S _32, it calculates the average of the pixel values in the block.
Then, the difference between the average value _{<S 11> - <R 11} >, <S 12> - <R 12>, <S 31> - <R 31>, <S 32> - calculating the value of _{<R 32>} Thus, the sign of the shift value α of each block is determined. In this case, for <S ₁₁ >-<R ₁₁ > and <S ₃₁ >-<R ₃₁ >, +2 is a positive number, and for <S ₁₂ >-<R ₁₂ >, <S ₃₂ >-<R ₃₂ > because it has a negative _-2, the S _{11, S 31} is embedded secret information _"1", the S _{12, S 32} it is detected that is embedded secret information "0".

As the shift value α is increased, the embedding strength (resistance) is increased. FIG. 12B shows S ₃₁ and S ₃₂ when the shift value α = + 1, −1 and when the shift value α = + 3, -3.
In the former, the average value difference <S ₃₁ > − <R ₃₁ > is +1, and in the latter, the average value difference <S ₃₁ > − <R ₃₁ > is +3. For example, as shown in FIG. 12B, when S ₃₁ is subjected to an attack that shifts a column of blocks by one and becomes S ′ ₃₁ , the average pixel value is 5.5 when the shift value α = ± 1. When the shift value α = ± 3, the average is 6.5. At this time, the average difference <S ′ ₃₁ > − <R ₃₁ > is +0.5, and in the latter, the average value difference <S ₃₁ > − <R ₃₁ > is +1.5, and the shift value α is large. The tolerance becomes stronger.
"A Digital Image Watermarking Scheme Withstanding StirMark Attack", Jiro Tanimoto, Akira Shiozaki, SCIS'99 The 1999 Symposium on Cryptography and Information Security Kobe, Japan, January 26-29,1999 The Institute of Electronics, Information and Communication Engineers

  As described above, in the digital watermark in which the secret information is embedded depending on whether the average of the pixel values in the block is larger or smaller than that of the original image, as the shift value α is increased, that is, in each block in the block. The tolerance can be increased as the value α uniformly applied to the pixels is increased.

However, as the shift value α is increased, the image degradation becomes conspicuous at the boundary between blocks. In FIG. 12, image degradation is not so conspicuous between blocks with large dispersion as indicated by R ₁₁ and R ₁₂ , but in particular, between blocks with small dispersion such as R ₃₁ and R ₃₂ , and one block In the case where the shift value α is added to the other block and the shift value α is subtracted from the other block, deterioration at the boundary between the blocks is large, and in some cases, a streak-like boundary becomes conspicuous.

  In the above description, the digital watermark is embedded by adding or subtracting the shift value α with respect to the average of the pixel values. However, the digital watermark is embedded in the conversion coefficient value used when performing some image processing on the pixel value. Similar problems arise in some cases.

  Therefore, the present invention blocks an image, and when embedding a digital watermark by shifting the average of pixel values or the average of transform coefficient values in the block, the present invention has the same level of resistance as the conventional embedding method and It is an object of the present invention to provide an electronic watermark embedding method, an electronic watermark embedding device, and an electronic watermark detection device that prevent deterioration of the boundary of the image (a streak-like boundary line).

The digital watermark information embedding device according to the present invention, which has been made to solve the above-mentioned problems, provides a secret to a pixel value which is luminance information and color information of each pixel constituting an image or a conversion coefficient value which performs an operation on the pixel value. An electronic watermark embedding method for embedding information, in which (a) an image or transform coefficient is divided into M rows × N columns (M and N are integers of 2 or more) and grouped into blocks or blocks (b) ) For a block in which secret information is embedded, a shift value α that changes the average of the pixel values or the average of transform coefficients in the block is set according to the secret information embedded in the block, and (c) the block after the secret information is embedded pixel value average or transform coefficient values mean inner (<A1 _ij>) as is interblock-pixel value average or transform coefficient values mean before embedding (<A0 _ij>) relationship of the formula (1), the block And so as to increase or decrease the uniform random number in accordance with the shift value α for each pixel value or each transform coefficient values of the inner.
<A1 _ij > = <A0 _ij > + α (1)
Here, <A1 _ij > is the average of pixel values or transform coefficient values in the block A1 _ij after embedding <A0 _ij > is the average of pixel values or transform coefficient values in the block A0 _ij before embedding i is the row number of the block , J is the block column number Shift value α is positive when the secret information is 1, negative when the secret information is 0 (or negative when the secret information is 1, and when the secret information is 0 Positive)

  That is, according to the present invention, when shifting the average of the pixel values and transform coefficient values in the block by the shift value α, the shift value α is not uniformly added to or subtracted from each pixel in the block. A uniform random number corresponding to the shift value α is obtained while the average of the pixel values is changed by the shift value α, and is added to or subtracted from each pixel in the block.

  According to the present invention, since each pixel in a block is added or subtracted as a uniform random number instead of a uniform value, even when secret information is embedded between blocks having a small variance, Pixel values vary, and the boundary between blocks can be made inconspicuous.

(Means and effects for solving other problems)
In addition, a digital watermark embedding method according to a second invention made to solve the above-described problem is a pixel value that is luminance information or color information of each pixel constituting an image, or a conversion coefficient value that performs an operation on the pixel value. In contrast, a digital watermark embedding method for embedding secret information, wherein an image or conversion coefficient is divided into M rows × N columns (M and N are integers of 2 or more) pixels or conversion coefficient groups, and is blocked. For the block (B _ij ) in which information is embedded, the average (<Ba _ij >, <Bb) for two different pixel values or two different transform coefficient values in the block according to the secret information embedded in the block _ij>) sets the shift value β as a reference when changing the said two kinds of pixel value average or converted absolute difference in the coefficient value average within the block after implantation secret information (| <Ba _ij> <Bb ij> _|) are uniform random number such that the relationship of formula (2), according to the pixel value or the converted coefficient values in the block to shift value beta.
| <Ba _ij > − <Bb _ij > | ≧ β (2)
Where <Ba _ij > is the average of the first pixel value or transform coefficient value <Bb _ij > is the average of the second pixel value or transform coefficient value Shift value β is <Ba when the positive number secret information is 1 _{When ij} > is greater than <Bb _ij > and the secret information is 0, it is small (or when the secret information is 1, <Ba _ij > is less than <Bb _ij > and when the secret information is 0, it is large. )

That is, according to the present invention, when the original image cannot be used at the time of detection, these two kinds of pixel values or averages of transform coefficient values (<Ba _ij >, <Bb _ij >) in the block are used. Embed confidential information according to the size relationship. At that time, since the size relationship is embedded as a uniform random number instead of a uniform value for each pixel in the block, even if the secret information is embedded between blocks with small variance, the pixel value at the boundary part The boundary line of the block can be made inconspicuous. The secret information “1” and “0” can be detected depending on which of the two types of pixel values or the average of the conversion coefficient values is larger.

In this case, it is preferable to use Cb and Cr among the three types of pixel values expressed in the (Y, Cb, Cr) color system as the two types of pixel values.
Since the human eye has low visibility with respect to the color components, when the original image cannot be used, it can be made inconspicuous by using Cb and Cr having large influence of the color components as pixel values for embedding. .

In the above invention, the uniform random number is a random number composed of an integer group of −U to V (0 <U <V, U and V are integers), and the relationship of Expression (3) is established with respect to the shift value α. It is preferable to set as follows.
α = (− U + V) / 2 (3)
According to the present invention, the range -U to V of the uniform random number value embedded in each pixel can be mechanically determined using the equation (3).

In addition, the digital watermark embedding device of the present invention, which has been made from another point of view, provides secret information with respect to a pixel value that is luminance information and color information of each pixel constituting an image or a conversion coefficient value that performs an operation on the pixel value. An electronic watermark embedding apparatus for embedding, wherein a block forming unit that blocks an image or transform coefficient into pixels divided by M rows × N columns (M and N are integers of 2 or more) or transform coefficients, and blocks A shift value setting unit in which a shift value α for changing the average of the pixel values or the average of the conversion coefficient values is set, and the average of the pixel values or the conversion coefficient values in the block after embedding the secret information based on the shift value α Each pixel value or transform coefficient value in the block is set so that (<A1 _ij >) has the relationship of the pixel value average or transform coefficient value average (<A0 _ij >) in the block before embedding and Expression (1). On the other hand, A uniform random number corresponding to the foot value α is calculated, and a uniform random number adding unit for adding to each pixel in the block is provided.

According to the present invention, the blocking unit divides an image or transform coefficients into blocks by dividing them into pixel groups or transform coefficient groups of M rows × N columns (M and N are integers of 2 or more). At this time, the number of rows and the number of columns of the block is not particularly limited as long as it is 2 or more. However, as the number of rows and the number of columns increases, the durability increases, but the amount of embedded information (number of bits) decreases. Set the appropriate number of rows and columns. The shift value setting unit sets a shift value α that changes the average of the pixel values or the average of the transform coefficient values in the block. The shift value α may be set to a predetermined value in advance, or may be changed by dispersion of pixel values for each block (for example, α is small when the variance is small and α is large when the variance is large). It may be set. Then, based on the shift value α, the uniform random number adding unit calculates the average pixel value or average conversion coefficient value (<A1 _ij >) in the block after embedding the secret information as the average pixel value or conversion factor in the block before embedding. A uniform random number corresponding to the shift value α is calculated for each pixel value or transform coefficient value in the block so that the relationship between the numerical average (<A0 _ij >) and the expression (1) is obtained. It is added to each pixel.

  According to the present invention, the digital watermark embedding apparatus embeds the shift value α as a uniform random number, instead of uniformly adding or subtracting it to each pixel in the block. Even when information is embedded, pixel values at the boundary portions vary, and the boundary lines of the blocks can be made inconspicuous.

  The digital watermark detection apparatus according to another aspect of the present invention, which is made to detect a digital watermark created by the digital watermark embedding apparatus, includes a secret information about an embedded image in which secret information is embedded and an original image. A block forming unit that is divided into block units each consisting of a pixel group or transform coefficient group of the same M rows × N columns (M and N are integers of 2 or more), and each block in which secret information in the embedded image is embedded And a block average value calculation unit that calculates an average of pixel values or an average of transform coefficient values in the block as a block average value for each block of the original image corresponding to the block, and the calculated block corresponding to each other A secret information extraction unit that extracts secret information for each block based on the magnitude relationship between the block average values of the embedded image and the original image; It has to.

According to the present invention, the blocking unit of the digital watermark detection apparatus uses the same M rows × N columns (M and N are two or more) as those when the secret information is embedded in the embedded image and the original image in which the secret information is embedded. Are divided into block units composed of pixel groups or conversion coefficient groups. The block average value calculation unit calculates the average of the pixel values or the average of the transform coefficient values in each block for each block in which the confidential information is embedded in the embedded image and each block of the corresponding original image. Calculate as an average value. The secret information extraction unit extracts secret information for each block based on the calculated magnitude relationship between the block average values of the embedded image and the original image corresponding to each other.
Thereby, the embedded secret information can be extracted from the embedded image and the original image using a uniform random number.

Further, another digital watermark embedding device of the present invention, which has been made from another point of view, provides a secret to a pixel value that is luminance information and color information of each pixel constituting an image or a conversion coefficient value that performs an operation on the pixel value. A digital watermark embedding device for embedding information, wherein the image or transform coefficient is divided into M rows × N columns (M and N are integers of 2 or more) pixel groups or transform coefficient groups, and a blocking unit; For the block (B _ij ) in which the secret information is embedded, the averages (<Ba _ij >, <<) of two different pixel values or two different transform coefficient values in the block according to the secret information embedded in the block. A shift value setting unit in which a shift value β serving as a reference when changing Bb _ij >) is set, and an absolute value of the average of the two types of pixel values or the average of transform coefficient values in the block after embedding the secret information A uniform random number corresponding to the shift value β is applied to each pixel value or each transform coefficient value in the block so that the difference (| <Ba _ij > − <Bb _ij > |) is in the relationship of the expression (2). A uniform random number adding unit that calculates and converts two types of pixels or conversion coefficients into uniform random numbers is provided.
| <Ba _ij > − <Bb _ij > | ≧ β (2)
Where <Ba _ij > is the average of the first pixel value or transform coefficient value <Bb _ij > is the average of the second pixel value or transform coefficient value Shift value β is <Ba when the positive number secret information is 1 _{When ij} > is greater than <Bb _ij > and the secret information is 0, it is small (or when the secret information is 1, <Ba _ij > is less than <Bb _ij > and when the secret information is 0, it is large. )

According to the present invention, the blocking unit divides an image or transform coefficient into blocks of pixels divided into M rows × N columns (M and N are integers of 2 or more). The shift value setting unit, for the block in which the secret information is embedded, for each average or two different types of transform coefficient values for two different pixel values or transform coefficient values in the block according to the secret information embedded in the block A shift value β is set as a reference when shifting the average of (<Ba _ij >, <Bb _ij >). The shift value β may be set to a predetermined value in advance, or may be set to change depending on the dispersion of pixel values for each block. Then, the uniform random number adding unit, based on the shift value β, calculates the absolute difference (| <Ba _ij > − <Bb _ij > |) between the average of two types of pixel values or the average of transform coefficient values in the block after embedding the secret information. ) In the equation (2), a uniform random number corresponding to the shift value β is calculated for each pixel value or each conversion coefficient value in the block, and the two types of pixels or conversion coefficients are determined as a uniform random number. To.

As a result, the digital watermark embedding apparatus embeds the shift value β as a uniform random number rather than uniformly adding or subtracting it to each pixel in the block, so that the secret information is embedded between the blocks having a small variance. Even at this time, the difference in pixel values at the boundary portion varies, and the boundary of the block can be made inconspicuous. Then, it is detected that the secret information is 1 or 0 depending on whether <Ba _ij > is larger or smaller than <Bb _ij >.

  In addition, the digital watermark detection apparatus for detecting the digital watermark created by the digital watermark embedding apparatus has the same M rows × N columns (when the secret information is embedded) with respect to the embedded image in which the secret information is embedded. M and N are integers greater than or equal to 2) Blocking unit that divides into block units consisting of pixel groups or transform coefficient groups, and two different pixel values in the block for each block in which the confidential information in the embedded image is embedded Block average difference calculation unit for calculating the average difference of the transform coefficients or the average difference of the transform coefficient values, and the secret information for extracting the secret information for each block based on the magnitude relationship of the calculated block average difference And an extraction unit.

According to the present invention, the blocking unit of the digital watermark detection apparatus has the same M rows × N columns (M and N are integers of 2 or more) as in the case of embedding the secret information for the embedded image in which the secret information is embedded. The block is divided into pixel units or transform coefficient groups. The block average value calculation unit calculates, as the block average value, an average difference between two different types of pixel values or an average difference between transform coefficient values for each block in which secret information is embedded. The secret information extraction unit extracts secret information for each block based on the calculated magnitude relationship between the two types of block averages.
Thereby, the secret information embedded using the uniform random number can be extracted.

(Embodiment 1)
Hereinafter, embodiments of the present invention will be described with reference to the drawings. Here, an embodiment in which an original image can be used when detecting secret information from an embedded image will be described. FIG. 1 is a block diagram showing a configuration of a digital watermark embedding apparatus 10 according to an embodiment of the present invention.
The digital watermark embedding device 10 stores software including a program for executing processing for embedding secret information, a CPU for executing arithmetic processing by executing the program, control parameters necessary for execution of the program, and calculation processing. The computer device includes a memory serving as a work area, image data, control parameters, and hardware including an input / output device that inputs and outputs commands.

  In order to explain specific processes and functions executed by the apparatus, the apparatus configuration is described separately for each functional block. The digital watermark information embedding apparatus 10 includes a blocking unit 11, a shift value setting unit 12, It consists of a uniform random number adding unit 13.

  The blocking unit 11 performs processing for dividing the image into small blocks based on the original image data. The block unit size may be changed according to the amount of information to be embedded and the embedding strength (resistance). Specifically, the number of rows and the number of columns are selected to appropriate values. Here, the block unit is preset to 4 rows × 4 columns.

  The shift value setting unit 12 sets a shift value α used when embedding secret information. That is, in order to embed the secret information, a process of changing the average of the pixel values in the block by the shift value α is performed. Before the process is performed, the magnitude of the shift value α is set in advance. Specifically, the shift value α may be initialized as a constant in advance, or an appropriate shift value α may be given by numerical input from the input device. The number of rows and columns in units of blocks may be calculated and set from the relationship between the number of rows and columns of the entire image and the amount of information to be embedded (number of bits). In the present embodiment, it is assumed that 2 is set as the shift value α.

Based on the set shift value α, the uniform random number adding unit 13 calculates the pixel value average (<A1 _ij >) in the block after embedding the secret information as the pixel value average (<A0 _ij >) in the block before embedding. ) And the expression (1), a uniform random number corresponding to the shift value α is calculated for each pixel value in the block, and an operation for adding to each pixel in the block is performed.

Next, the operation of the digital watermark embedding device 10 will be described. FIG. 2 is a flowchart showing an example of an operation procedure performed by the digital watermark embedding apparatus 10. FIG. 3 is a diagram for explaining the state of pixel values in which secret information is embedded.
Here, a case where the embedded image A1 is created from the original image A0 will be described. First, the original image A0 (reference image) is divided into blocks of 4 rows × 4 columns (S1001). i the row direction of the block number, the j and the column direction of the block number, each block of the reference image A0 _{A0 ij,} each block of the embedded image A1 is expressed as _{A1 ij.} Then, all or part of the divided blocks are selected as embedded blocks in which secret information is embedded (S1002). Here, in FIG. 3, the blocks A0 ₁₁ and A0 ₁₂ are the blocks to be embedded. The blocks A0 ₃₁ and A0 ₃₂ are blocks in which secret information is embedded by a conventional method for comparison. For the sake of convenience, the block having the smallest variance is selected for convenience (all pixel values are a constant value 5), and the boundary line is easily noticeable when the secret information is embedded.

  Subsequently, the shift value α is set. Here, since the shift value α is set in advance and stored in the memory, the value (= 2) is read (S1003).

Subsequently, a uniform random number that satisfies Equation (1) is calculated.
<A1 _ij > ≧ <A0 _ij > + α (1)
Here, <A1 _ij > is the average of the pixel values in the block A1 _ij after embedding <A0 _ij > is the average of the pixel values in the block A0 _ij before embedding i is the row number of the block, j is the column number of the block The value α is positive when the secret information is 1, and negative when the secret information is 0.

That is, the average of the pixel values in the blocks A1 ₁₁ and A1 ₁₂ after embedding changes by +2 (relative to the block A0 ₁₁ before embedding) or changes by -2 (relative to the block A0 ₁₂ before embedding). A uniform random number combination is calculated. In the present embodiment, since 16 pixel groups are included per block (4 rows × 4 columns), a combination of 16 random numbers is calculated for each of A1 ₁₁ and A1 ₁₂ .
The uniform random number can take various combinations. For example, it is a random number consisting of an integer group of −U to V (0 <U <V, U and V are integers), and the relationship of Expression (3) may be established with respect to the shift value α.
α = (− U + V) / 2 (3)
Specifically, when α is set to 2, by setting the maximum random number V to 5, −U is determined to be −1 from the equation (3). Therefore, the numbers {-1, 0, 1, 2, 3, 4, 5} are selected as uniform random numbers, and these numbers are distributed to 16 pixels with a uniform probability. If the uniform random maximum value V is set to a different value, different sets of uniform random numbers can be obtained.

Subsequently, the calculated 16 uniform random numbers are added to or subtracted from each pixel (S1005).
The embedded image shown in FIG. 3 shows the state of the pixel value at this time. A1 ₁₁ is - uniform random number {1,0,1,2,3,4,5} is added to _{A0 11.} Further, _{A1 12} are - uniform random number {1,0,1,2,3,4,5} is subtracted from _{A0 12.}
Thus, instead of uniformly adding or subtracting the shift value α to each pixel value, a uniform random number corresponding to the shift value α is obtained while the average of the pixel values is changed by the shift value α. Thus, it is adjusted to each pixel in the block.
As a result, even if the secret information for increasing the average value is embedded in the block A0 ₁₁ and the block A1 ₁₂ having a small variance, and the secret information for decreasing the average value is embedded in the other block A1 the value at the boundary between the ₁₁ and the block 11 ₁₂ is randomly boundary is less likely to be perceived.

Incidentally, for comparison, in blocks A0 ₃₁ and A0 ₃₂ in which each pixel value is uniformly changed by a shift value ± 2 using the conventional embedding method, a step having a large pixel value is formed in a line at the boundary portion. This will be perceived as a streak.

  Next, a digital watermark detection apparatus will be described. FIG. 4 is a block diagram showing a configuration of the digital watermark information detection apparatus 30 according to the embodiment of the present invention. The digital watermark detection apparatus 30 includes a computer device similar to the digital watermark embedding apparatus 10. Although the digital watermark embedding device 10 and the digital watermark detection device 30 are described as separate devices, they may be incorporated into one device so that both embedding and detection of secret information can be performed. .

  In order to explain specific processes and functions executed by the present apparatus, it will be described separately for each functional block. The digital watermark information detecting apparatus 30 includes a blocking unit 31, a block average value calculating unit 32, and secret information extraction. Part 33.

  The blocking unit 31 performs a process for dividing the original image data and the embedded image into units of small blocks. The size of this block unit is divided into the same block unit as when the secret information is embedded by the embedding device 10. Here, the block unit is 4 rows × 4 columns.

The block average value calculation unit 32 calculates the average of the pixel values in the block as the block average value for each of the block in which the confidential information is embedded in the embedded image and the corresponding block of the original image.
The secret information extraction unit 33 extracts the secret information for each block based on the calculated magnitude relationship between the block average values of the embedded image and the original image corresponding to each other. Specifically, the block average values in the blocks corresponding to each other are compared between the embedded image and the original image, and when the average of the pixel values of the embedded image is larger than the average of the pixel values of the original image, the secret information “1 "Is embedded. When the average pixel value of the embedded image is smaller than the average pixel value of the original image, it is determined that the secret information “0” is embedded. When both are the same, it is determined that they are not embedded. Thus, it is extracted whether the secret information of the block is “1” or “0”.

FIG. 5 is a flowchart showing an operation example of the digital watermark information detection apparatus 30. Here, a case where secret information is extracted from the original image A0 and the embedded image A1 in FIG. 3 will be described. First, the original image A0 and the embedded image A1 are divided into block units of 4 rows × 4 columns (S1101). Subsequently, for each block of the original image and the embedded image, block average values of pixel values are calculated in order from A0 ₁₁ and A1 ₁₁ respectively (S1102).
Subsequently, the difference between the block average values of the corresponding original image and the embedded image is calculated, and the size is compared (S1103). As a result, the value of the secret information “1” or “0” embedded in each block is extracted (S1104).
The above processing is performed for all blocks, and all secret information is extracted.
Thereby, digital watermark information can be detected.

(Example)
Next, in order to confirm the effect of the present invention, an example of an image when a uniform image is divided into blocks and 1-bit secret information (luminance information) is embedded in each block by the following method will be shown.

6A and 6C, the pixel value of each block is increased or decreased by a uniform random number when increasing or decreasing the pixel value average. In (b) and (d), when the pixel value average is increased or decreased, the pixel value of each block is increased or decreased uniformly. All of (a) to (d) have the same pixel value average.
When (a) and (b) are compared and (c) and (d) are compared, the boundaries between the blocks are less conspicuous in (a) and (c).

Also, the secret information pattern embedded in the block (the luminance increase / decrease pattern in the block) is a pattern in which (a) and (c) are different, and (b) and (d) are different in pattern (a ) And (b) and (c) and (d) have the same pattern.
If the pixel value is increased or decreased uniformly, bring two images with different secret information patterns into the same image, and take the difference between the images to find out what size block It can be estimated whether the secret information is embedded or whether the bits (secret information) embedded in the two blocks are the same or different. For example, FIG. 6F is image data obtained by extracting the difference between (b) and (d), and the block unit can be easily estimated. If such information is estimated, there is a risk that the embedded watermark (secret information) may be deleted or rewritten.

  On the other hand, FIG. 6E shows image data obtained by extracting the difference between (a) and (c). As shown in (a) and (c), when secret information is embedded using a uniform random number, even if the difference between these two images is taken, it is possible to hide what size block the secret information is embedded in. it can. Therefore, safety is increased.

(Embodiment 2)
Next, a second embodiment of the present invention will be described with reference to the drawings. Here, an embodiment in which the original image cannot be used when detecting the secret information from the embedded image will be described. That is, an example will be described in which only the embedded image is used for detection, two types of pixel values included in the embedded image are obtained, and the secret information is detected from the average magnitude relationship of these. Specifically, the secret information is embedded using two types of color information of Cb and Cr of the (Y, Cb, Cr) color system as two types of pixel values. The reason why Cb and Cr are used is that deterioration is less conspicuous for human eyes than luminance information Y when secret information is embedded.

FIG. 7 is a block diagram showing a configuration of the digital watermark embedding apparatus 20 according to the embodiment of the present invention.
The digital watermark embedding device 20 stores software including a program for executing processing for embedding secret information, a CPU for executing arithmetic processing by executing this program, control parameters necessary for execution of the program, and calculation processing. The computer device includes a memory serving as a work area, image data, control parameters, and hardware including an input / output device that inputs and outputs commands.

  In order to explain specific processes and functions executed by the apparatus, the apparatus configuration is described separately for each functional block. The digital watermark information embedding apparatus 20 includes a blocking unit 21, a shift value setting unit 22, It consists of a uniform random number adding unit 23.

  The blocking unit 21 performs processing for dividing the image into small blocks based on the original image data. The block unit size (number of pixels) may be changed according to the amount of information to be embedded and the embedding strength (resistance). Here, the block unit is preset to 4 rows × 4 columns.

The shift value setting unit 22 changes the average of pixel values of pixel values Cb and Cr of two different types of color information in a block in accordance with the secret information embedded in the block for the block in which the secret information is embedded. A shift value β used as a reference for the time is set.
This shift value β is a value that is the lower limit of the difference between the average values of the two types of pixel values Cb and Cr. Note that the larger β is, the greater the resistance is, but the boundary is more conspicuous. Here, 2 is set as β.
In the following description, a block in which secret information is embedded will be described in a generalized expression as B _ij , a pixel value for Cb as Ba _ij , and a pixel value for Cr as Bb _ij . That is, instead of the color components Cb and Cr, a description will be given in a generalized form so that the present embodiment can be used by combining other pixel values, and the present invention can be applied as it is when other pixel values are used. To do.

The uniform random number adding unit 23 determines that the absolute difference (| <Ba _ij > − <Bb _ij > |) of the average of two types of pixel values or the average of the conversion coefficient values in the block after embedding the secret information is a relation of the formula (2). as a, each pixel value in the block, and calculates a uniform random number in accordance with the shift value beta, performs calculation to replace two of the pixel values Ba ij of block B _ij, the Bb _ij uniform random numbers . At this time, when the secret information is 1, the pixel value average <Ba _ij > is larger than the pixel value average <Bb _ij > by a shift value β (or more), and conversely when the secret information is 0 become.

Next, rules for embedding secret information will be described with reference to FIG. As shown in FIG. 8A, when the secret information embedded in the block B _ij is “1”, the average <Ba _ij > on the pixel value side of the color information Cb is equal to the average pixel value <Bb _ij > of Cr. Is made larger by β (= 2). When the secret information embedded in the block B _ij is “0”, the pixel value average <Ba _ij > of the color information Cb is set to be smaller by β than the Cr pixel value average <Bb _ij >. Then, from the beginning, the current pixel value is used as it is for the block that has such a magnitude relationship, and the block within the block is forcibly established for the block that does not have this magnitude relationship. Change the pixel value.

More specifically, as shown in FIG. 8B, the blocks B ₁₁ and B ₁₂ both have a small variance of the pixel values of the pixel components Ba and Bb (the pixel values are constant), and the pixel component Ba (Cb). Is 7 and 3 for the pixel component Bb (Cr).

Assume that the secret information “0” is embedded in B ₁₁ and the secret information “1” is embedded in B ₁₂ . If, in the case of uniformly changing the pixel values, since the 2 as β is set, for example, for the pixel component Ba of the block B ₁₁ 6, so that 8 is embedded for the pixel component Bb. (Pixel component Ba is uniformly subtracted by β / 2 to 6 and pixel component Bb is uniformly added by β / 2 to 8). Similarly, for the pixel component Ba of the block _{B 12} is 8, so that the 6 is embedded for the pixel component Bb. As a result, for B ₁₁ , the pixel value average <Ba ₁₁ > is 2 smaller than the pixel value average <Bb ₁₁ >, and the secret information “0” is buried. On the other hand, with respect to B ₁₂ , <Ba ₁₂ > is 2 larger than <Bb ₁₂ >, and the secret information “1” is embedded.

In this case, the boundary between the block B ₁₁ and the block B ₁₂ is stepped pixel value occurs in a row, so streaky boundary becomes conspicuous.
Therefore, the pixel values in the block are not uniformly changed, but the pixel values are given by uniform random numbers while satisfying the relationship of absolute difference | <Ba _ij > − <Bb _ij > | ≧ β. In the above example, component Ba of the block _{B 11,} the component of the block _{B 12} Ba, components of the block _{B 12} Ba, a pixel value of a component Bb of the block _{B 12,} as in the example shown in FIG. 8 (c) uniform Use random numbers. That is, the pixel value average for component Bb components Ba and block _{B 12} of the block _{B 11 <Ba 11>, <} Bb 12> is next 6, the pixel values for the component Bb components Ba and block _{B 11} of the block _{B 12} While the average <Ba ₁₂ > and <Bb ₁₁ > are 8, the respective pixel values are made uniform random numbers.

Various uniform random numbers can be selected. For example, the pixel component Ba is uniformly added or subtracted by β / 2, and the pixel component Bb is uniformly subtracted or added by β / 2. An absolute difference of β is provided between the pixel components Ba and Bb. After _{obtaining the} average pixel values <Ba _ij > and <Bb _ij > of the block at this time, 16 uniform values are obtained so that the average value is equal to or larger than the <Ba _ij > and <Bb _ij >. A random number may be selected.

Next, an operation procedure example using the above-described embedding rule by the digital watermark embedding apparatus 20 will be described. FIG. 9 is a flowchart showing an operation example of the digital watermark embedding apparatus 20. The embedding operation will be described with reference to FIG. 8 again.
First, as shown in FIG. 8B, the image B is divided into blocks of 4 rows × 4 columns (S2001). The order of blocks in which the secret information is embedded is determined in advance, and the secret information is embedded in order according to the information amount (number of bits). Here, the first blocks B ₁₁ and B ₁₂ are the blocks to be embedded. B ₁₁ and B ₁₂ are blocks with the smallest variance, and the boundaries are the most conspicuous blocks when the secret information is embedded.

  Subsequently, the shift value β is set (S2002). Since the shift value β is set to 2 in advance and stored in the memory, the value (= 2) is read out.

Subsequently, the absolute difference (| <Ba _ij > − <Bb _ij > |) between the two types of pixel value averages or transform coefficient value averages in the block after embedding the secret information is in the relationship of Equation (2): A uniform random number corresponding to the shift value β is calculated for each pixel value in the block (S2003).
| <Ba _ij > − <Bb _ij > | ≧ β (2)
Where <Ba _ij > is the average of the first pixel values <Bb _ij > is the average of the second pixel values Shift value β is <Ba _ij > from <Bb _ij > when the positive number secret information is 1 Large and small when secret information is 0 (or large when secret information is 1 and <Ba _ij > is smaller than <Bb _ij > and secret information is 0)

That is, a combination of numbers satisfying the expression (2) and the two types of pixel values in the blocks B ₁₁ and B ₁₂ are respectively uniform random numbers is calculated. In the present embodiment, 16 pixel groups are included in one block (4 rows × 4 columns), so that each of the component Ba and the component Bb of the block B _{11 and} the component Ba and the component Bb of the block B ₁₂ A combination of random numbers is calculated.

Subsequently, the calculated 16 uniform random numbers are assigned to each pixel of each block. (S2004).
The embedded image shown in FIG. 8C shows the state of the pixel value at this time. Ba of B ₁₁ is assigned a uniform random number of {3, 4, 5, 6, 7, 8, 9} so that the pixel value average <Ba ₁₁ > is 6. Further, Bb of B ₁₂ is assigned a uniform random number of {5, 6, 7, 8, 9, 10, 11} so that the pixel value average <Bb ₁₂ > is 8.

In this way, instead of changing each pixel value uniformly, a uniform random number is obtained and applied to each pixel in the block while changing the average difference of the pixel values by the shift value β.
As a result, the boundary value is random even between the block B ₁₁ and the block B ₁₂ having a small variance, and the boundary is hardly perceived.

  Next, a digital watermark detection apparatus will be described. FIG. 10 is a block diagram showing a configuration of a digital watermark information detection apparatus 40 according to an embodiment of the present invention. The digital watermark detection device 40 is composed of the same computer device as the digital watermark embedding device 20. The digital watermark embedding device 20 and the digital watermark detection device 40 may be incorporated into one device so that both the embedding and detection of secret information can be performed.

  In order to explain specific processes and functions executed by the apparatus, the explanation will be made separately for each functional block. The digital watermark detection apparatus 40 includes a blocking unit 41, a block average value calculating unit 42, a secret information extracting unit. 43.

  The blocking unit 41 performs processing for dividing the embedded image into units for each small block. The size of this block unit is divided into the same block unit as when the secret information is embedded by the embedding device 20. Here, the block unit is 4 rows × 4 columns.

The block average value calculation unit 42 calculates, for each block in the embedded image, the average <Ba _ij > and <Bb _ij > of the two types of pixel values Ba and Bb in the block as the block average value.
The secret information extraction unit 43 extracts the secret information for each block based on the calculated magnitude relationship between the two types of block average values. Specifically, “1” is detected as secret information when <Ba _ij > is larger than <Bb _ij >, and “0” is detected when it is smaller.

FIG. 11 is a flowchart illustrating an operation example of the digital watermark detection apparatus 30. First, the image is divided into block units of 4 rows × 4 columns (S2101). Subsequently, for each block of the embedded image, block average values <Ba _ij > and <Bb _ij > of two types of pixel values Ba and Bb are calculated (S2102).
Subsequently, a difference between the two types of block average values is calculated, and a size comparison is performed (S2103). According to the result, the value of the secret information “1” or “0” embedded in each block is extracted (S2104).
The above processing is performed for all blocks, and all secret information is extracted.
In this way, it is possible to embed or detect a digital watermark in which the boundary between blocks is not noticeable.

  In the first and second embodiments, the image data is divided into blocks and the secret information is embedded. However, the image data is divided into blocks using the conversion coefficient used when compression conversion is performed, and the secret information is embedded. But you can. Also in this case, embedding and detection can be performed in the same manner as in the case of an image.

  The present invention can be used in a digital watermark embedding device that embeds confidential information as an electronic watermark in an image, and a digital watermark detection device that detects an embedded digital watermark.

1 is a block diagram showing a configuration of a digital watermark embedding device 10 according to an embodiment of the present invention. 6 is a flowchart showing an operation example of the digital watermark embedding apparatus 10. The figure explaining the state of the pixel value which embedded secret information. The block diagram which shows the structure of the digital watermark information detection apparatus 30 which is one Embodiment of this invention. 5 is a flowchart showing an operation example by the digital watermark information detection apparatus 30. The figure which shows the example of an image which embedded confidential information. The figure which shows the example of an image which embedded confidential information. The figure which shows the example of an image which embedded confidential information. The block diagram which shows the structure of the digital watermark embedding apparatus 20 which is one Embodiment of this invention. FIG. 5 is a diagram showing the principle of embedding processing by the digital watermark embedding device 20. 5 is a flowchart showing an operation example by the digital watermark embedding apparatus 20. The block diagram which shows the structure of the digital watermark information detection apparatus 40 which is one Embodiment of this invention. 5 is a flowchart showing an operation example by the digital watermark detection apparatus 30. The figure explaining the principle of the method of embedding secret information and the conventional method of detecting secret information using the average of the pixel value for every block.

Explanation of symbols

10, 20: Digital watermark embedding device 11, 21: Blocking unit 12, 22: Shift value setting unit 13, 23: Uniform random number adding unit 30, 40: Digital watermark detecting device 31, 41: Blocking unit 32, 42: Block average value calculation unit 33, 43: Secret information extraction unit

Claims (8)

An electronic watermark embedding method for embedding secret information in a pixel value that is luminance information and color information of each pixel constituting an image or a conversion coefficient value that performs an operation on the pixel value,
The image or conversion coefficient is divided into M rows × N columns (M and N are integers of 2 or more) and divided into blocks to form blocks.
For a block (A _ij ) that embeds secret information, a shift value α that changes the average of pixel values or the average of transform coefficient values in the block according to the secret information embedded in the block is set.
The pixel value average or transform coefficient value average (<A1 _ij >) in the block after embedding the secret information is the relationship between the pixel value average or transform coefficient value average (<A0 _ij >) in the block before embedding and the relationship of Equation (1): The digital watermark embedding method, wherein the uniform random number corresponding to the shift value α is increased or decreased for each pixel value or each transform coefficient value in the block.
<A1 _ij > ≧ <A0 _ij > + α (1)
Here, <A1 _ij > is the average of pixel values or transform coefficient values in the block A1 _ij after embedding <A0 _ij > is the average of pixel values or transform coefficient values in the block A0 _ij before embedding i is the row number of the block , J is the block column number Shift value α is positive when the secret information is 1, negative when the secret information is 0 (or negative when the secret information is 1, and when the secret information is 0 Positive) An electronic watermark embedding method for embedding secret information in a pixel value that is luminance information and color information of each pixel constituting an image or a conversion coefficient value that performs an operation on the pixel value,
The image or conversion coefficient is divided into M rows × N columns (M and N are integers of 2 or more) and divided into blocks to form blocks.
For the block (B _ij ) in which the secret information is embedded, the averages (<Ba _ij >, <<) of two different pixel values or two different transform coefficient values in the block according to the secret information embedded in the block. Set a shift value β as a reference when changing Bb _ij >),
The block so that the absolute difference (| <Ba _ij > − <Bb _ij > |) of the average of the two types of pixel values or the average of the transform coefficient values in the block after embedding the secret information is in the relationship of Expression (2). A digital watermark embedding method characterized in that each pixel value or each conversion coefficient value is a uniform random number corresponding to the shift value β.
| <Ba _ij > − <Bb _ij > | ≧ β (2)
Where <Ba _ij > is the average of the first pixel value or transform coefficient value <Bb _ij > is the average of the second pixel value or transform coefficient value Shift value β is <Ba when the positive number secret information is 1 _{When ij} > is greater than <Bb _ij > and the secret information is 0, it is small (or when the secret information is 1, <Ba _ij > is less than <Bb _ij > and when the secret information is 0, it is large. ) 3. The digital watermark embedding method according to claim 2, wherein Cb and Cr are used among the three types of pixel values expressed in the (Y, Cb, Cr) color system as the two types of pixel values. The uniform random number is a random number composed of an integer group of −U to V (0 <U <V, U and V are positive integers), and the relationship of Expression (3) is established with respect to the shift value α. 2. The digital watermark embedding method according to 1.
α = (− U + V) / 2 (3) An electronic watermark embedding device that embeds secret information into a pixel value that is luminance information and color information of each pixel constituting an image or a conversion coefficient value that performs an operation on the pixel value,
A blocking unit that blocks an image or transform coefficient into a pixel group or transform coefficient group of M rows × N columns (M and N are integers of 2 or more);
A shift value setting unit in which a shift value α for changing the average of the pixel values in the block or the average of the transform coefficient values is set;
Based on the shift value α, the pixel value average or transform coefficient value average (<A1 _ij >) in the block after embedding the secret information is equal to the pixel value average or transform coefficient value average (<A0 _ij >) in the block before embedding. A uniform random number corresponding to the shift value α is calculated for each pixel value or each transform coefficient value in the block so as to satisfy the relationship of Expression (1), and is added to each pixel in the block. An electronic watermark embedding apparatus comprising a random number adding unit.
<A1 _ij > ≧ <A0 _ij > + α (1)
Where <A1 _ij > is the average of pixel values or transform coefficient values in the block after embedding <A0 _ij > is the average of pixel values or transform coefficient values in the block before embedding i is the block row number, and j is the block The column number shift value α is positive when the secret information is 1, and negative when the secret information is 0 (or negative when the secret information is 1, and positive when the secret information is 0). An electronic watermark detection apparatus for detecting secret information embedded by the electronic watermark embedding apparatus according to claim 5,
The embedded image and the original image in which the secret information is embedded are divided into block units each including a pixel group or a transform coefficient group of the same M rows × N columns (M and N are integers of 2 or more) as when the secret information is embedded. A blocking unit;
A block average that calculates the average of pixel values or the average of transform coefficient values in each block as a block average value for each block in which embedded confidential information is embedded in the embedded image and each block of the original image corresponding to the block. A value calculator,
A digital watermark detection apparatus, comprising: a secret information extraction unit that extracts secret information for each block based on a calculated magnitude relationship between a block average value of an embedded image and an original image corresponding to each other. An electronic watermark embedding device that embeds secret information into a pixel value that is luminance information and color information of each pixel constituting an image or a conversion coefficient value that performs an operation on the pixel value,
A blocking unit that blocks an image or transform coefficient into a pixel group or transform coefficient group of M rows × N columns (M and N are integers of 2 or more);
For the block (B _ij ) in which the secret information is embedded, the averages (<Ba _ij >, <<) of two different pixel values or two different transform coefficient values in the block according to the secret information embedded in the block. A shift value setting unit in which a shift value β serving as a reference when changing Bb _ij >) is set;
The block so that the absolute difference (| <Ba _ij > − <Bb _ij > |) of the average of the two types of pixel values or the average of the transform coefficient values in the block after embedding the secret information is in the relationship of Expression (2). A uniform random number adding unit that calculates a uniform random number corresponding to the shift value β for each pixel value or each conversion coefficient value in the pixel value and converts the two types of pixels or conversion coefficients into a uniform random number. A digital watermark embedding device as a feature.
| <Ba _ij > − <Bb _ij > | ≧ β (2)
Where <Ba _ij > is the average of the first pixel value or transform coefficient value <Bb _ij > is the average of the second pixel value or transform coefficient value Shift value β is <Ba when the positive number secret information is 1 _{When ij} > is greater than <Bb _ij > and the secret information is 0, it is small (or when the secret information is 1, <Ba _ij > is less than <Bb _ij > and when the secret information is 0, it is large. ) A digital watermark detection apparatus for detecting secret information embedded by the digital watermark embedding apparatus according to claim 7,
A block forming unit that divides the embedded image in which the secret information is embedded into blocks each including a pixel group or transform coefficient group of the same M rows × N columns (M and N are integers of 2 or more) when the secret information is embedded; ,
A block average difference calculation unit that calculates an average difference between two different pixel values in a block or an average difference between transform coefficient values as a block average difference for each block in which secret information in an embedded image is embedded;
An electronic watermark detection apparatus comprising: a secret information extraction unit that extracts secret information for each block based on the calculated block average difference.