JP2013009162A - Electronic watermark image embedding device and electronic watermark image detection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve an electronic watermark embedding processing for guaranteeing calculation of a same unique ID from a plurality of images where different pieces of electronic watermark information are embedded respectively.SOLUTION: Random signal sequence calculation means 11 generates random signal sequences of the number as the same as the number of bits of the watermark information bit data. Correlation value calculation means 12 calculates correlation values between input images and the random signal sequences. Watermark signal coefficient determination means 13 determines coefficients for the correlation values by inputting threshold information and watermark information bit-data. Watermark signal addition means 14 adds values obtained by multiplying the coefficients determined by the watermark signal coefficient determination means 13 to the random signal sequences to pixel values of input images as watermark signals and outputs them as watermark embedded images.

Description

  The present invention relates to an image digital watermark embedding device that embeds a digital watermark in an image, and an image digital watermark detection device that detects watermark information and image ID information from the embedded image.

As a conventional image digital watermark embedding apparatus, for example, Patent Document 1 discloses a technique of embedding a hash value of digital information (content) as a digital watermark.
Non-Patent Document 1 discloses a technique for generating a bit called a visual hash as a unique ID for an image based on the sign of the correlation between the original image and a plurality of random Gaussian signals.

Japanese Patent No. 3854804

Fridrich, Jiri, “Visual Hash For Oblivious Watermarking”, Proc. SPIE Vol. 3971, p.286-294, Security and Watermarking of Multimedia Contents II, Ping Wah Wong; Edward J. Delp; Eds.

However, in the conventional digital watermark embedding device disclosed in Patent Document 1, since the hash value is calculated from the image signal, the image signal is changed even slightly (it is a minute change that cannot be detected by human eyes). There is a problem that hash information changes even in the case of a change in signal value that is not generally considered to have been changed.
As a method for solving this problem, there is a bit generation method disclosed in Non-Patent Document 1. In this document, a method for calculating an image-specific ID that is not easily affected by a minute image change is shown. If this method is adopted as the hash value calculation unit of Patent Document 1, the case where the hash value is embedded in the image itself and managed is solved.

  However, in a system in which hash information is not embedded in the image itself, but is centrally managed by the image publisher, a hash value is generated from an image in which other information independent of the image is embedded as a digital watermark (for example, a digital watermark) In such a case, another problem occurs.

  That is, when hash calculation is performed on an image in which other information independent of the image is embedded as a digital watermark in advance, the unique ID of the image can be calculated using the hash bit generation method shown in Non-Patent Document 1. The same unique ID cannot always be calculated for a plurality of images embedded with different digital watermark information. In general, the operation of embedding a digital watermark in an image and distinguishing individual contents while retaining unique ID information in the image is contradictory, and the same unique ID than an image embedded with a separate digital watermark. The probability that information can be extracted is low. Therefore, when the hash information is centrally managed by the image issuer, it is necessary to construct a database that associates the embedded digital watermark information or the image embedded with the digital watermark with the individual hash information. There was a problem that the operation cost increased. In this regard, if it is guaranteed that constant hash information can always be calculated from the same image signal without depending on the embedded watermark information, the database range is the same as the original image signal and the single hash. It will be limited only to the association with information, which is extremely advantageous in terms of construction and operation costs.

  The present invention has been made to solve the above-described problems, and is a digital watermark embedding process that guarantees that the same unique ID is calculated from a plurality of images in which different digital watermark information is embedded. It is an object of the present invention to obtain an image digital watermark embedding device capable of realizing the above and an image digital watermark detection device for detecting the device.

  An image digital watermark embedding device according to the present invention comprises: random signal sequence calculation means for generating a random signal sequence having the same number as the number of bits of watermark information bit data based on key information; and a correlation value between an input image and a random signal sequence A correlation value calculating means for calculating a watermark signal coefficient, a watermark signal coefficient determining means for determining a coefficient for adding a watermark signal, using threshold value information and watermark information bit data set in advance as to the correlation value, and a random signal There is provided watermark signal adding means for embedding a digital watermark by adding a value obtained by multiplying the series by the coefficient determined by the watermark signal coefficient determining means to the pixel value of the input image as a watermark signal.

  The image digital watermark embedding device according to the present invention determines a coefficient by inputting threshold information and watermark information bit data for a correlation value between an input image and a random signal sequence, and multiplies the random signal sequence by the coefficient. Is added to the pixel value of the input image as a watermark signal, so that it is possible to ensure that the same unique ID is calculated from a plurality of images in which different digital watermark information is embedded.

It is a block diagram which shows the image digital watermark embedding apparatus by Embodiment 1 of this invention. It is explanatory drawing which shows operation | movement of the random signal sequence calculation means of the image digital watermark embedding apparatus by Embodiment 1 of this invention. Correlation value C ₁ of the image digital watermark embedding apparatus according to the first embodiment of the _{invention, C 2, C 3, ...} , is an explanatory diagram showing the occurrence probability distribution of C _n. It is explanatory drawing for setting a correlation value outside the range of-(alpha)-+ (alpha) with respect to the watermark embedding image of the image digital watermark embedding apparatus by Embodiment 1 of this invention. BRIEF DESCRIPTION OF THE DRAWINGS It is a block diagram of the image digital watermark detection apparatus by Embodiment 1 of this invention. It is explanatory drawing which shows the threshold value information of the image digital watermark embedding apparatus by Embodiment 2 of this invention. It is explanatory drawing of the watermark signal coefficient of the image digital watermark embedding apparatus by Embodiment 2 of this invention. It is a block diagram of the image digital watermark embedding apparatus by Embodiment 3 of this invention. It is explanatory drawing which shows the concept of the operation | movement of the bit conversion of the image digital watermark embedding apparatus by Embodiment 3 of this invention. It is a block diagram which shows the image digital watermark detection apparatus by Embodiment 3 of this invention.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing an image digital watermark embedding apparatus according to Embodiment 1 of the present invention. The image digital watermark embedding apparatus 10 shown in FIG. 1 includes a random signal sequence calculation unit 11, a correlation value calculation unit 12, a watermark signal coefficient determination unit 13, and a watermark signal addition unit 14.

  The random signal sequence calculation means 11 is a means for generating a random signal sequence having the same number as the number of watermark information bits by inputting specific key information and the number of watermark information bits. The correlation value calculation means 12 is a means for calculating correlation values between the original image to be embedded with the watermark and the plurality of random signal sequences calculated by the random signal sequence calculation means 11. The watermark signal coefficient determining means 13 is a coefficient for adding the random signal sequence by the watermark signal adding means 14 at the subsequent stage based on the correlation value calculated by the correlation value calculating means 12 and receiving the watermark information bit data and threshold information. Is a means of calculating The watermark signal adding unit 14 embeds a watermark by adding a signal sequence obtained by multiplying a plurality of random signal sequences calculated by the random signal sequence calculating unit 11 to the coefficient calculated by the watermark signal coefficient determining unit 13 to the original image. This is a means for generating a completed output image.

Next, the operation of the image digital watermark embedding apparatus 10 according to the first embodiment will be described.
The image digital watermark embedding apparatus 10 uses the original image O, key information K, watermark information bit data W, and threshold information α to be embedded as input information.
First, in the random signal sequence calculation means 11, the key information K and the watermark information bit data W are input with the number of watermark information bits (the input line is omitted in the figure) and the image size of the original image O as inputs. A random signal sequence composed of the same number of random signals as the number of pixels of the image size of O is calculated by the same number as the number of watermark information bits.

FIG. 2 shows the specific processing. The random signal sequence calculation means 11 generates random values by the number of [image size of the original image O] × [number of watermark information bits (assumed to be n bits)] using the input key information K. In the figure, the random values are collected for each image size of the original image O and expressed as “Random Image” as R ₁ , R ₂ , R ₃ ,..., R _n . And the process by LPF (Low Pass Filter) is performed with respect to those random value sets. This is expressed as “Random Smooth Patterns” in the drawing as P ₁ , P ₂ , P ₃ ,... P _n . The random signal sequence calculation means 11 outputs P ₁ , P ₂ , P ₃ ,... P _n as a random signal sequence.

Next, the correlation value calculation means 12 calculates correlation values between the original image O and the random signal sequences P ₁ , P ₂ , P ₃ ,... P _n . The correlation value between the original image O and the random signal sequence P ₁ is C ₁ , the correlation value between the original image O and the random signal sequence P ₂ is C ₂ , and the correlation value between the original image O and the random signal sequence P ₃ is C _3. , a correlation value between ... original image O and the random signal sequence _{P n} is expressed as _{C n,} the correlation value calculating means 12 _{C 1, C 2, C 3} , and outputs the ... _{C n.}

Next, the watermark signal coefficient determining means 13 uses the correlation values C ₁ , C ₂ , C ₃ ,... C _n calculated by the correlation value calculating means 12, and the watermark signal adding means 14 at the subsequent stage uses the original image O. To the random signal sequence P ₁ , P ₂ , P ₃ ,... P _n calculated by the random signal sequence calculation means.

Here, a method of embedding a digital watermark will be described.
FIG. 3 shows the concept of the probability distribution of correlation values C ₁ , C ₂ , C ₃ ,... C _n . Usually, since the random signal sequence is an image that is almost unrelated to the original image O, the correlation value is concentrated in the vicinity of 0, and conversely, almost no occurrence is seen in the vicinity of ± 1. . Therefore, a value of ± α (0 <α <1) is set as the correlation threshold value, and the probability of occurrence of an event outside the threshold value ± α (an event in which the absolute value of the correlation value is greater than or equal to α) is almost zero. (For example, 10 to the power of −8). This ± α is “threshold information”.

As an example of how to embed digital watermark bit information,
When bit 1 is embedded in the i-th bit, the correlation between P _i and the watermark embedded image is set outside the range of −α to + α.
When bit 0 is embedded in the i-th bit, the correlation between P _i and the watermark embedded image is set within the range of −α to + α.
The method can be considered. That is, the idea is to distinguish the watermark bit information when the image correlation with the random signal sequence P _i is in the range (−α, + α) and outside the range (of course, the above example and bit 1). / 0 may be interchanged, and bit 1 may be set within the range of -α to + α, and bit 0 may be set outside the range of -α to + α).

To set to be a correlation value outside the range of -α~ + α with respect to the watermark embedding image, performs addition of the random signal sequence P _i in the original image O. In general, when the random signal sequence P _i is added to the original image O, the correlation value increases. Therefore, when the original correlation value C _i is a positive value and bit 1 is embedded in the i-th bit, the correlation value is greater than or equal to + α. Therefore, the random signal series _Pi may be added. This is shown in FIG. Note that the example of FIG. 4 is a case where the original correlation value C _i is a positive value. Conversely, if the original correlation value C _i is a negative value, if bit 1 is embedded in the i-th bit, the correlation value is Must be subtracted rather than added to make.

However, only when they simply the sum of P _i itself to the original image O (or subtracted), typically one which can greatly exceed the threshold + alpha, watermarking image to the original of the original image O There is a concern that it will deteriorate greatly. Therefore, the random signal sequence P _i is not added as it is to the original image O, but a value obtained by multiplying a specific coefficient value m _i (0 <m _i <1) is added. This is called a “watermark signal coefficient”. In the watermark signal coefficient determining unit 13 to compute the watermark signal coefficients m _i and main processing.

An example of the watermark signal coefficient calculation process is as follows.
When the pixel value of the original image O is x _j and the pixel value of the random signal sequence P _i is y _j (j is a pixel number), the correlation value r _xy between the input signal A and the random signal sequence P _i is Expressed as an expression.

ここで、

として、式１を変形すると、次式のように表現される。

Here, as shown in FIG. 4, when calculating the random signal sequence P _i, a correlation value between the image signal obtained by adding to the original image O of the random signal sequence P _i obtained by multiplying the watermark signal coefficients m _i, the correlation The value r _xy ′ is as follows:

here,

When Expression 1 is transformed, it is expressed as the following expression.

を満たす必要がある。この式を変形すると、

となる。この式はｍ_ｉに関する連立二次不等式と見ることができる。この不等式を解くと、

となる（｜α｜＜１、かつ相関値の定義より（ｖ_ｘｖ_ｙ−ｖ_ｘｙ ^２）≧０が常に成り立つため、上記不等式解の２つの境界値は、どちらも常に実数値を取ることになる）。 In order for the correlation value thus obtained to be outside the range (−α, + α), the square value needs to exceed α ² . That is,

It is necessary to satisfy. If this equation is transformed,

It becomes. This equation can be seen as a simultaneous secondary inequalities related to m _i. Solving this inequality,

Become (| α | <1, and the definition of the correlation values ^{_{(v x v y -v xy 2}} ) ≧ 0 for always holds, two boundary values of the inequality solutions, both to always take the real value become).

Therefore, to embed the bit 1, in order to the outside of -α~ + α a correlation between P _i and the watermark embedding images, the m _i is set to a range satisfying the above inequality, to embed the bit 0 , in order to in the range of -α~ + α a correlation between P _i and the watermark embedding image, set the m _i in the range that does not meet the above inequality (i.e., sandwiched between the boundary value of the inequality solutions range setting) may be (Note that in the case that may take m i _{= 0} is as much as possible and m _{i =} 0 in order to avoid image quality degradation). Incidentally, if positive is the original correlation value r _xy, the m _i under the range of the condition to a positive value, setting the m _i a negative value if it is negative.

In this way, it is possible to determine the watermark signal coefficients m _i to allocate the r _{xy 'in} a range out of -α~ + α in accordance with the embedded bit information while maintaining the same reference numerals and r _xy . In the watermark signal coefficient determining unit 13, for each bit value of the watermark information bit data W, the original image each pixel value of O, random signal sequence calculating unit 11 each pixel value of the random signal sequence P _i output from, and the threshold value using information alpha, and outputs the calculated watermark signal coefficients m _i for each bit by the method described above.

Next, the watermark signal adding means 14, with m _i output by the watermark signal coefficient determining unit 13, multiplying the m _i to each pixel value of the random signal sequence P _i output from the random signal sequence calculating unit 11 Then, the multiplication result is added to each corresponding pixel of the original image O. Then, the resulting image E is output as a watermark embedded image. The watermark embedded image E is an output image of the image digital watermark embedding device 10.

Next, detection of digital watermark will be described.
FIG. 5 is a block diagram of an image digital watermark detection apparatus according to the present invention. The image digital watermark detection apparatus 20 of FIG. 5 includes a random signal sequence calculation unit 21, a correlation value calculation unit 22, a watermark information calculation unit 23, and a unique ID calculation unit 24.

  The random signal sequence calculating means 21 is a means for generating a random signal sequence having the same number as the number of watermark information bits by inputting specific key information and the number of watermark information bits. The correlation value calculation unit 22 is a unit that calculates correlation values between the original image that is the target of watermark detection and the plurality of random signal sequences calculated by the random signal sequence calculation unit 21. The watermark information calculation unit 23 is a unit that calculates bit information embedded as a watermark based on the correlation value and threshold information calculated by the correlation value calculation unit 22. The unique ID calculation unit 24 is a unit that calculates ID information unique to an image based on the correlation value calculated by the correlation value calculation unit 22.

Next, the operation of the image digital watermark detection apparatus 20 will be described.
The image digital watermark detection apparatus 20 uses the image E, key information K, watermark information bit number n, and threshold information α to be detected as the input information.
First, in the random signal sequence calculation means 21, a random signal composed of random signals of the same number as the number of pixels of the image size of the detection target image E, with the key information K, the watermark information bit number n, and the image size of the detection target image E as inputs. The sequence is calculated by the same number as the number of watermark information bits. The specific processing is the same as that of the random signal sequence calculation means 11 of the image digital watermark embedding apparatus 10 and is as shown in FIG.

Next, the correlation value calculation means 22 calculates correlation values between the detection target image E and the random signal sequences P ₁ , P ₂ , P ₃ ,... P _n . The correlation value between the detection target image E and the random signal sequence P ₁ is C ₁ , the correlation value between the detection target image E and the random signal sequence P ₂ is C ₂ , and the correlation value between the detection target image E and the random signal sequence P _3. the _C 3, a correlation value between ... detection target image E and the random signal sequence _{P n} is expressed as _{C n,} the correlation value calculating means _{22 C 1, C 2, C} 3, outputs ... _{C n} To do.

Next, in the watermark information calculation means 23, the correlation values C ₁ , C ₂ , C ₃ ,... C _n calculated by the correlation value calculation means 22 and the threshold information α input to the image digital watermark detection apparatus 20 are used. Using this, the watermark information bit data W embedded as a watermark in the detection target image E is calculated. Specifically, as described above, how to embed digital watermark bit information is:
When embedding bit 1 in the i-th bit, the correlation value is set outside the range of -α to + α.
When embedding bit 0 in the i-th bit, the correlation value is set within the range of −α to + α.
Since the watermark information calculation means 23 uses the threshold information α,
When C _i is outside the range of −α to + α, it is determined that bit 1 is embedded in the i-th bit.
When C _i is in the range of −α to + α, it is determined that bit 0 is embedded in the i-th bit.
In this manner, each bit information is detected. The detection results are combined and output as watermark information bit data W. The watermark information bit data W is output information of the image digital watermark detection apparatus 20.

On the other hand, the unique ID calculating unit 24 calculates ID information unique to the image using the correlation values C ₁ , C ₂ , C ₃ ,... C _n calculated by the correlation value calculating unit 22. Specifically, the correlation values C ₁ , C ₂ , C ₃ ,... C _n are examined to determine whether “bit 1 is positive” or “bit 0 if negative” (and vice versa). It may be determined that “0 is positive if it is positive” and “1 if it is negative”. The calculation results are combined and output as image unique ID information I. The image unique ID information I becomes output information of the image digital watermark detection apparatus 20.

  The digital watermark embedding method in the image digital watermark embedding apparatus 10 is calculated in this way because the digital watermark is embedded so as not to move the positive / negative of the correlation value between the original input image and the random signal sequence. The image unique ID in principle matches the image unique ID of the original image before embedding the watermark.

  As described above, according to the image digital watermark embedding device of the first embodiment, the random signal sequence calculating means for generating the same number of random signal sequences as the number of bits of the watermark information bit data based on the key information, and the input Correlation value calculation means for calculating a correlation value between an image and a random signal sequence; and input threshold value information and watermark information bit data set in advance with respect to the correlation value, a coefficient for watermark signal addition is determined. A watermark signal coefficient determining unit; and a watermark signal adding unit that embeds a digital watermark by adding a value obtained by multiplying a random signal sequence by a coefficient determined by the watermark signal coefficient determining unit to a pixel value of an input image as a watermark signal. Therefore, a method for calculating the unique ID based on the sign of the correlation value with the random signal sequence for the original image is accompanied by a change in the bit value, etc. Embedding the digital watermark information without affecting an effect that is possible.

  Also, according to the image digital watermark embedding device of the first embodiment, the watermark signal coefficient determining means receives ± α (0 <α <1) as the value of the threshold information, and if bit 1 is embedded, the correlation absolute value If the bit 0 is embedded so that the correlation absolute value is larger than α when the bit 0 is embedded, or if the bit 0 is embedded, the bit is determined so that the correlation absolute value is smaller than α. In the case of embedding 1, since the coefficient is determined so that the correlation absolute value becomes larger than α, the digital watermark information can be embedded without affecting the bit value.

  Further, according to the image digital watermark detection apparatus of the first embodiment, based on the key information, random signal sequence calculation means for generating the same number of random signal sequences as the number of bits of the watermark information, the input image, the random signal sequence, A correlation value calculation means for calculating the correlation value, a watermark information calculation means for calculating watermark information bit data by inputting threshold information preset for the correlation value, and bit information based on the sign of the correlation value A unique ID calculating means for determining and calculating the unique ID of the image, so that the correlation value with the random signal sequence is calculated for the image in which the digital watermark is embedded, and the image uniqueness is determined by looking at its positive / negative While calculating the ID, information embedded as a digital watermark can be calculated in parallel.

  Also, according to the image digital watermark detection apparatus of the first embodiment, the watermark information calculation means receives ± α (0 <α <1) as the value of the threshold information, and if the correlation absolute value is smaller than α, a bit 1. If the correlation absolute value is larger than α, it is determined to be bit 0, or if the correlation absolute value is smaller than α, it is determined to be bit 0, and if the correlation absolute value is larger than α, it is determined to be bit 0. Information embedded as a digital watermark can be reliably detected.

Embodiment 2. FIG.
In the first embodiment, the threshold information α is set and the correlation value with the random signal sequence is distributed to the inside and outside of the range (−α, + α) to embed the digital watermark bit information. In the second embodiment, an example in which digital watermark information is prevented from being detected from an image with no watermark embedded by setting a plurality of threshold information and embedding the digital watermark will be described.

  When detection is performed using an image signal with no digital watermark embedded as an input image in accordance with the digital watermark detection method described in the first embodiment, the occurrence of a correlation value with a random signal sequence is shown as a conceptual diagram in FIG. As you can see, most of them fall within the range (-α, + α). Therefore, as the detection result of the digital watermark, it is determined that “bit 0 is embedded in all bits” in almost all cases. That is, a phenomenon that some bit information can be obtained from an image that should not be embedded (not only in the case of all bits 0) (whether this phenomenon causes a problem in actual use) Depending on how it is used).

The purpose of this embodiment is to avoid this phenomenon. The configuration of the image digital watermark embedding apparatus according to the second embodiment is the same as that shown in FIG. 1 and will be described with reference to FIG.
FIG. 6 shows a number line of correlation values when two kinds of threshold information α1 and α2 are set as input of the image digital watermark embedding apparatus 10. Note that the value of α ₁ is set so that the occurrence probability of an event outside the threshold ± α ₁ (an event in which the absolute value of the correlation value is equal to or greater than α ₁ ) is approximately 0 (for example, 10 −8). .

Here, the watermark signal coefficient determination means 13 makes the following determination regarding the relationship between the embedding of the watermark information bits 0/1 and the correlation value.
· ~-Α _2: embed a bit 1.
-Α ₂ to -α ₁ : Bit 0 is embedded.
-Α ₁ to + α ₁ : No bits are embedded.
+ Α ₁ to + α ₂ : Bit 0 is embedded.
+ Α ₂ ˜: Bit 1 is embedded.
Therefore, each threshold information of alpha _1, fitting the alpha ₂ in equation 2 described in Embodiment 1, those corresponding to the inequality boundary value (alpha ₁ for the watermark signal coefficients _{m i γ 1+, γ 1-,} α 2 gamma ₂₊ those corresponding _to, after seeking gamma _2-to), to determine the value of the watermark signal coefficients m _i in segmentation as shown in FIG.

Next, the image digital watermark detection apparatus 20 according to the second embodiment will be described. The configuration of the image digital watermark detection apparatus 20 according to the second embodiment is the same as that shown in FIG. 5 and will be described with reference to FIG.
In the watermark information calculation means 23 of the image digital watermark detection apparatus 20 according to the second embodiment, the correlation values C ₁ , C ₂ , C ₃ ,... C _n calculated by the correlation value calculation means 22 and the image digital watermark detection apparatus. Using the plurality of threshold information α ₁ and α ₂ input to 20, watermark information bit data W embedded as a watermark in the detection target image E is calculated. Specifically, the value of _{C i} is,
-˜α ₂ : It is determined that bit 1 is embedded.
-Α ₂ to -α ₁ : It is determined that bit 0 is embedded.
-Α ₁ to + α ₁ : It is determined that no bits are embedded.
+ Α ₁ to + α ₂ : It is determined that bit 0 is embedded.
+ Α ₂ ˜: It is determined that bit 1 is embedded.
In this way, the correlation value corresponding to each bit of the watermark information bit data W is determined. Here, if even one bit of each bit is “determined that no bit is embedded”, it is determined that the entire watermark information bit data W is invalid, and digital watermark information is embedded in the corresponding image. Judge that it is not.

  As described above, according to the image digital watermark embedding device of the second embodiment, the watermark signal coefficient determination unit receives ± α1 and ± α2 (0 <α1 <α2 <1) as threshold information values, and When embedding bit 1, the coefficient is determined so that the correlation absolute value is larger than α1 and smaller than α2, and when bit 0 is embedded, the coefficient is determined so that the correlation absolute value is larger than α2, or bit 0 is set Since the coefficient is determined so that the correlation absolute value is larger than α1 and smaller than α2 when embedding, and the correlation absolute value is larger than α2 when embedding bit 1, the correlation value without embedding bits. It is possible to prevent digital watermark information bit data from being obtained from an image that should not have anything embedded therein.

  Also, according to the image digital watermark detection apparatus of the second embodiment, the watermark information calculation means receives ± α1 and ± α2 (0 <α1 <α2 <1) as threshold information values, and the correlation absolute value is α1. Bit 1 if greater and less than α2, bit 0 if the correlation absolute value is greater than α2, or bit 0 if the correlation absolute value is greater than α1 and less than α2, bit 0, correlation absolute value When is larger than α2, since it is determined to be bit 1, digital watermark information can be obtained with certainty.

Embodiment 3 FIG.
In the second embodiment, an example in which a digital watermark is prevented from being detected from an image in which no watermark is embedded by setting a plurality of threshold information and embedding the digital watermark has been described. An embodiment will be described as a third embodiment in which information is error-corrected and bit conversion is performed so that watermark signal bit data in which all 0 bits are out of the correction range of the error-correction encoded bit data.

FIG. 8 is a configuration diagram of the image digital watermark embedding apparatus 10a shown in the present embodiment. The image digital watermark embedding device 10a includes a random signal sequence calculation unit 11a, a correlation value calculation unit 12, a watermark signal coefficient determination unit 13a, a watermark signal addition unit 14, an error correction encoding unit 15, and a bit conversion unit 16. Correlation value calculation means 12 and watermark signal addition means 14 are the same means as in the first and second embodiments. The error correction coding means 15 is means for performing error correction coding with the watermark information bit data as input, and the bit conversion means 16 is means for bit-converting the error correction coded bit data. This is a configuration unique to mode 3. Further, the random signal sequence calculation means 11a of the third embodiment is not the number of watermark information bits (n bits) but the number of redundant coding bits (n + k bits) of the error correction coding method used in the error correction coding means 15. Only random signal sequences P ₁ , P ₂ , P ₃ ,... P _{n + k} are output. Further, the watermark signal coefficient determination means 13a is configured to calculate a coefficient using the watermark signal bit data output from the bit conversion means 16 instead of the watermark information bit number (n bits).

Next, the operation of the third embodiment will be described. The basic operations of the random signal sequence calculating unit 11a, the correlation value calculating unit 12, the watermark signal coefficient determining unit 13a, and the watermark signal adding unit 14 are described in the embodiment. Since it is the same as 1 and 2, explanation here is omitted. However, the random signal sequence calculation means 11a is not the number of watermark information bits (n bits) but the random signal sequence P corresponding to the number of redundant coding bits (n + k bits) of the error correction coding method used in the error correction coding means 15. ₁ , P ₂ , P ₃ ,... P _{n + k} are output. It is assumed that the number of redundant coding bits is preset in accordance with the error correction coding method of the error correction coding means 15.

  The error correction coding means 15 receives the watermark information bit data W and performs redundant coding (BCH code or the like) for error correction. The purpose of the redundant coding is to improve the detection capability by error correction and to create a state in which “digital watermark is not detected” outside the correction capability of this redundant encoded bit string by making the “non-digital watermark detection” state possible. This is to prevent detection of significant digital watermark information. The bit conversion means 16 performs bit conversion for this purpose.

FIG. 9 shows the concept of these operations on the code space. When the number of bits of the watermark information bit data W is ⁿ , the watermark information bit data W is expressed as one point (element) on the 2 ⁿ bit code space. If this is converted into (n + k) -bit redundant encoded bit data R by error correction encoding, the redundant encoded bit data R is expressed as one point (element) in the 2 ^{n + k-} bit code space. Further, if the bit correction capability of the error correction code is r, R is a range of radius r centered on a point on the 2 ^{n + k} bit code space (the number of different bits is less than r bits compared to R 2 ⁽ a set of ^{n + k} bit code data) is defined as an error correctable range.

Here, for example, when an error correction coding system such as a BCH code is used, if the information bit data is all 0s, the redundant coded bit data is also a bit string of (n + k) bits composed of all 0s. On the other hand, when digital watermark detection is performed from an image in which no digital watermark is embedded, the probability distribution of the correlation values with the random signal sequences P ₁ , P ₂ , P ₃ ,... P _{n + k} is illustrated in FIG. As shown in the figure, since most of them are concentrated in the vicinity of 0, when bit 0/1 is determined by the threshold value ± α, most of the bits are determined to be bit 0.

  For this reason, if the redundantly encoded bit data of “all 0” is embedded in the image as it is as the digital watermark signal bit data, the detection result when the digital watermark is detected from the image in which no digital watermark is embedded. In the end, it is necessary to determine that an image in which no digital watermark is embedded is “all bits 0 are embedded”. In order to prevent such a situation, it is considered to shift the state where “all bits 0” is detected as an “abnormal state” and shift it outside the range of the correction capability of the error correction code.

  Bit conversion means 16 performs bit conversion for performing XOR operation on redundant encoded bit data R generated from watermark information bit data W with redundant encoded bit data R and arbitrarily determined fixed bit data having the same number of bits. The watermark signal is finally output as watermark signal bit data Q to be embedded in an image as a watermark signal. Thus, when the watermark information bit data W is “all bits 0”, the watermark signal bit data Q is not “all bits 0”, and conversely, “all bits 0” is detected as the watermark signal bit data. Can be placed outside the error correction coding correction capability (depending on how to set an arbitrary fixed bit string for performing the XOR operation). For this reason, it is possible to clearly distinguish from the case where the watermark information bit data of “all bits 0” is embedded with the “state in which nothing is embedded” being “error correction impossible”.

  Needless to say, the error correction coding used by the error correction coding means 15 is not limited to the BCH code, and the bit conversion performed by the bit conversion means 16 is also limited to the XOR operation with arbitrary fixed bit data. (However, the error encoding method used in the error correction encoding means 15 is not required for the subsequent bit conversion in the case where the encoded bit string of “all bits 0” is outside the error correction capability in the first place.) Becomes).

Next, an image digital watermark detection apparatus according to Embodiment 3 will be described.
FIG. 10 is a configuration diagram of the image digital watermark detection apparatus 20a of the present embodiment. The image digital watermark detection apparatus 20a in FIG. 10 includes a random signal sequence calculation unit 21a, a correlation value calculation unit 22, a watermark information calculation unit 23a, a unique ID calculation unit 24, a bit conversion unit 25, and an error correction decoding unit 26. . Here, the correlation value calculation means 22 and the unique ID calculation means 24 have the same configuration as that of the first and second embodiments shown in FIG. Bit conversion means 25 is means for converting watermark signal bit data calculated by the watermark information calculation means 23a, and error correction decoding means 26 is means for performing error correction decoding processing using the data output from the bit conversion means 25 as input. These means have a configuration unique to the present embodiment. Further, the random signal sequence calculating means 21a is configured to generate the same number of random signal sequences as the number of bits of the watermark signal bit data Q of the watermark information calculating means 23a, and the watermark information calculating means 23a is based on the threshold information α. The watermark signal bit data Q is calculated.

  The bit conversion means 25 performs a bit conversion means corresponding to the bit conversion processing in the bit conversion means 16 in the image digital watermark embedding apparatus 10 of FIG. 8 on the watermark signal bit data Q output from the watermark information calculation means 23a, and performs redundancy. Output as encoded bit data R. Specifically, as with the bit conversion means 16, bit conversion is performed to perform an XOR operation between redundantly encoded bit data and arbitrarily determined fixed bit data having the same number of bits. The error correction decoding unit 26 uses the error correction encoding method (for example, BCH code) of the error correction encoding unit 15 in the image digital watermark embedding apparatus 10a of FIG. 8 for the redundant encoded bit data R output from the bit conversion unit 25. Is output as watermark information bit data W.

  As described above, according to the image digital watermark embedding device of the third embodiment, the error correction coding means for outputting the redundant encoded bit data subjected to the redundant error correction encoding with the watermark information bit data as input, and the redundant Bit conversion means for performing bit conversion with the encoded bit data as input and outputting watermark signal bit data. The random signal sequence calculation means is a redundancy in the error correction coding means in place of the number of bits of the watermark information bit data. A random signal sequence having the same number of bits as the number of encoded bits is generated, and the watermark signal coefficient determination unit receives the watermark signal bit data output from the bit conversion unit instead of the watermark information bit data as a watermark. Since the coefficients for signal addition are determined, the watermark signal bit data is all bit 0. There it is possible to prevent that this by performing range become as bit conversion of the encoded bit sequences set, nothing embedded electronic watermark embedding bit information from the supposed image not obtained.

  Further, according to the image digital watermark detection apparatus of the third embodiment, the random signal sequence calculation means for generating the same number of random signal sequences as the number of bits of the watermark signal bit data based on the key information, the input image and the random signal Correlation value calculation means for calculating a correlation value with a sequence, watermark information calculation means for calculating watermark signal bit data using threshold value information preset for the correlation value, and bit input using watermark signal bit data Since it has bit conversion means for performing conversion and outputting redundant encoded bit data, and error correction decoding means for performing error correction decoding using redundant encoded bit data as input and outputting watermark information bit data as a decoding result It is possible to prevent the digital watermark embedding bit information from being obtained from an image that should not be embedded.

  In the present invention, within the scope of the invention, any combination of the embodiments, or any modification of any component in each embodiment, or omission of any component in each embodiment is possible. .

  10, 10a Image digital watermark embedding device, 11, 11a, 21, 21a Random signal sequence calculation means, 12, 22 Correlation value calculation means, 13, 13a Watermark signal coefficient determination means, 14 Watermark signal addition means, 15 Error correction coding Means, 16-bit conversion means, 20, 20a image digital watermark detection device, 23, 23a watermark information calculation means, 24 unique ID calculation means, 25-bit conversion means, 26 error correction decoding means.

Claims (8)

Random signal sequence calculation means for generating the same number of random signal sequences as the number of bits of watermark information bit data based on the key information;
Correlation value calculating means for calculating a correlation value between the input image and the random signal sequence;
Watermark signal coefficient determining means for determining a coefficient for adding a watermark signal by inputting threshold value information set in advance and the watermark information bit data with respect to the correlation value;
An image digital watermark comprising watermark signal adding means for embedding a digital watermark by adding a value obtained by multiplying the random signal sequence by the coefficient determined by the watermark signal coefficient determining means to a pixel value of the input image as a watermark signal Implanting device.   The watermark signal coefficient determination means receives ± α (0 <α <1) as the threshold information value. When bit 1 is embedded, the correlation absolute value is smaller than α, and when bit 0 is embedded, correlation absolute The coefficient is determined so that the value is larger than α, or when embedding bit 0, the coefficient is set so that the correlation absolute value becomes smaller than α, and when bit 1 is embedded, the coefficient so that the correlation absolute value becomes larger than α. The image digital watermark embedding apparatus according to claim 1, wherein:   The watermark signal coefficient determination means receives ± α1 and ± α2 (0 <α1 <α2 <1) as threshold information values, and when bit 1 is embedded, the correlation absolute value is larger than α1 and smaller than α2. In addition, when embedding bit 0, the coefficient is determined so that the correlation absolute value becomes larger than α2, or when bit 0 is embedded, the bit is set so that the correlation absolute value is larger than α1 and smaller than α2. 2. The image digital watermark embedding apparatus according to claim 1, wherein when 1 is embedded, the coefficient is determined so that the correlation absolute value is larger than α2. Error correction coding means for outputting redundant encoded bit data subjected to redundant error correction encoding with watermark information bit data as input; and
Bit conversion means for performing bit conversion using the redundant encoded bit data as input, and outputting watermark signal bit data; and
Random signal sequence calculation means, in place of the number of bits of the watermark information bit data, the number of redundant coding bits in the error correction coding means, to generate the same number of random signal sequences as the number of bits,
2. The watermark signal coefficient determining means determines a coefficient for adding a watermark signal by using the watermark signal bit data output from the bit converting means as an input instead of the watermark information bit data. The image digital watermark embedding apparatus according to any one of claims 1 to 3. Random signal sequence calculation means for generating the same number of random signal sequences as the number of bits of watermark information based on the key information;
Correlation value calculating means for calculating a correlation value between the input image and the random signal sequence;
Watermark information calculating means for calculating watermark information bit data using threshold information preset for the correlation value as input;
An image digital watermark detection apparatus comprising: unique ID calculation means for determining bit information based on the sign of a correlation value and calculating a unique ID of an image.   The watermark information calculation means receives ± α (0 <α <1) as the threshold information value, and determines bit 1 when the correlation absolute value is smaller than α and bit 0 when the correlation absolute value is larger than α. 6. The image digital watermark detection apparatus according to claim 5, wherein bit 0 is determined when the correlation absolute value is smaller than α, and bit 0 is determined when the correlation absolute value is larger than α.   The watermark information calculation means inputs ± α1 and ± α2 (0 <α1 <α2 <1) as threshold information values. If the correlation absolute value is larger than α1 and smaller than α2, bit 1 and the correlation absolute value are If greater than α2, bit 0 is determined, or if the correlation absolute value is greater than α1 and less than α2, bit 0 is determined, and if the correlation absolute value is greater than α2, bit 1 is determined. The image digital watermark detection apparatus according to claim 5. Random signal sequence calculation means for generating the same number of random signal sequences as the number of bits of watermark signal bit data based on the key information;
Correlation value calculating means for calculating a correlation value between the input image and the random signal sequence;
Watermark information calculating means for calculating the watermark signal bit data, using threshold information preset for the correlation value as input;
Bit conversion means for performing bit conversion using the watermark signal bit data as input, and outputting redundant encoded bit data;
An image digital watermark detection apparatus comprising: error correction decoding means for performing error correction decoding using the redundant encoded bit data as an input and outputting watermark information bit data as a decoding result.

Images