JP2001268339A - Method for embedding information in picture and method for extracting the same information - Google Patents

Abstract

PROBLEM TO BE SOLVED: To easily embed multi-bit information such as a digital signature in a small sized picture (graphic mark) using an electronic watermark. SOLUTION: An image is converted into frequency expression, and the value of the amplitude or phase of a cosine wave is displaced by bits to be embedded. That is, the coefficient values of the preliminarily decided elements of a DCT coefficient matrix 42 representing a picture block (8×8 pixels) 41 to be embedded out of a whole picture 43 are displaced by bits to be embedded. Also, not only the bits of information to be embedded but also the bits of the result of the rearray, conversion, or arithmetic operation of the bits can be obtained as the embed information to be displaced. Also, the bits to be displaced can be decided on the basis of the value of the amplitude and phase of another cosine waver or the value of the amplitude and phase of the cosine wave.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of embedding information in an image and a method of extracting information, and more particularly to a method of embedding information in a frequency-represented image by using a digital watermark and a method of extracting information.

[0002]

2. Description of the Related Art An image has a pixel expression and a frequency expression. In pixel representation, as shown in FIG.
The image is finely divided and R is set for each minimum block (pixel).
(Red) G (green) B (blue) luminance display or YUV display is performed. In the frequency expression, the brightness (brightness) and color of an image are expressed by superposition of cosine waves. Examples of the expression method include a method using a discrete cosine transform, a method using a Fourier transform, and a method using a wavelet transform. For example, in the method using the discrete cosine transform, as shown in FIG. 4, in order to represent an image by a two-dimensional wave, a section of 8 × 8 pixels is divided into a cosine wave in the X direction and a Y direction.
There is a frequency expression method expressed as a superposition of a cosine wave in a direction. The frequency expression by the discrete cosine transform in FIG. 4 is expressed by adding all 64 types of waveform weightings as in the following equation (1).

(Equation 1) Here, a _ij is a weight value, and is an amplitude value of the location in the image.

Conventionally, digital watermarking known as a technique for embedding information in an image includes a method of changing a pixel expression and a method of changing a frequency expression. In the latter, information is embedded by making a small change to the amplitude and phase of a cosine wave that is an element of an image. In the conventional digital watermarking technique, information such as an owner ID is embedded in a high-value image such as an art image in order to protect the copyright. For copyright protection, the emphasis has been placed on ensuring that information can be correctly detected even after image processing has been performed on an image in which information has been embedded.

A specific method is described in, for example, 1996 Symposium.
on Cryptography and Information Security, 31-G. In this method, several cosine waves are selected from a plurality of cosine waves that are elements of an image. FIG. 5 is an explanatory diagram of an information embedding method of a conventional digital watermark technique. In FIG.
5 illustrates a method of changing the weighting of only one portion in the image block illustrated in FIG. Weight value is 0 to 256
The open circles in the figure are multiple values of 16, and the black circles indicate intermediate values. As shown in FIG. 5, one-bit information of one cosine wave is embedded by changing the amplitude of the selected cosine wave. That is, assuming that the amplitude value of the cosine wave in the original image is V, if the information to be inserted is 1, V is replaced by a multiple of 16 closest to V. If the information to be inserted is 0, V is replaced with the closest to V (a multiple of 16 + 8). This process is performed on a plurality of cosine waves to embed a plurality of bits. In the figure, the amplitude value in the original image is 30 (00011
110), when the bit to be embedded is 1, the amplitude value is set to 32 (00100000), and when the bit to be embedded is 0, the amplitude value is set to 24 (0001).
1000).

When detecting insertion information from an image, the amplitude value is read for each of the plurality of changed cosine waves. Depending on whether the read value is close to a multiple of 16 or (multiple of 16 + 8), the bit value of the insertion information is set to 1 or 0.
Is determined. In this method, even if the information-embedded image is processed, 0 and 1 are not erroneously detected unless the amplitude of the cosine wave changes by 8 or more, so that the information can be correctly detected even from the image after the image processing. . That is, as long as the amplitude value does not change from the white circle in FIG. 5 to the value of the adjacent black circle in the processing, erroneous detection does not occur.

[0006] Applications of the digital watermark include authentication using a graphic mark in addition to the above-mentioned copyright protection. FIG. 6 is an explanatory diagram when a digital watermark is used in a conventional graphic mark authentication method, and FIG. 7 is an explanatory diagram when it is determined whether or not a graphic mark authenticated using a digital watermark is correct. As shown in FIG. 6, a digital signature 22 for a content 21 such as an arbitrary document is embedded in a graphic mark 23 (for example, an image of a face of a person to be authenticated) and used as an authentication mark 24. In addition to extracting the characteristic value of the graphic mark 23 (26), the characteristic value of the content 21 is also extracted (25), and a digital signature is created by calculation using the secret key 27 (28), and a digital watermark is embedded in the graphic mark 23. (29), and the authentication mark 24
Is combined with the content 21 (30). The certification mark 24 is issued as proof that the certification organization has been certified, and is combined (pasted) with the content 21 (30). By observing the authentication mark 24, the user of the content 21 knows that the accredited organization has accredited the content 21 for some reason.

Further, in order to confirm that the authentication mark 24 has not been falsified or forged, the digital signature 22 is extracted from the authentication mark 24 by the procedure shown in FIG. 7 and the signature is verified. That is, the characteristic value of the authentication mark 24 is extracted (26), the characteristic value of the content 21 is extracted (25), and after the digital watermark is extracted (29)
a) The digital signature is verified by performing an operation using the public key 27a (28a). As a result, if the determination 31 is YES, it is known that the authentication is correct, and if the determination 31 is NO, the authentication is incorrect. A digital watermark is used to embed the digital signature 22 in the authentication mark 24. In this application, it is not necessary to correctly detect the digital signature 22 from the authentication mark 24 after the image processing after the content 21 in which the authentication mark 24 is synthesized is subjected to the image processing. That is, the image processing resistance may be low.
This is because the fact that the digital signature (22) has been subjected to the image processing itself is a kind of tampering, and there is no problem if the verification of the digital signature (22) fails and it is determined that the digital signature (22) is tampered or forged. That is, since the content 21 before the authentication has been authenticated, if the graphic mark 24 is deformed, it cannot be authenticated as a matter of course. On the other hand, the owner ID and the like are at most about 10 bits, but the digital signature 22 has an information amount of about 100 to 2000 bytes, and it is necessary to embed many bits in the digital watermark of the authentication mark 24. .

[0008]

As described above, in the digital watermark for authentication, it is not necessary to endure image processing as compared with the conventional digital watermark, but it is necessary to embed a lot of information. Therefore, it is necessary to embed a lot of information in a short time by a simple operation.

It is an object of the present invention to solve these conventional problems and to provide a method of embedding information into an image and a method of extracting a large amount of information easily. Another object of the present invention is to provide a method of embedding information into an image which can easily embed multi-bit information such as a digital signature into a small-sized image (graphic mark) by an electronic watermark. It is in.

[0010]

In order to achieve the above object, a method of embedding information in an image according to the present invention is a method of embedding a predetermined bit of an amplitude value of a cosine wave forming an image into a bit of information to be embedded or embedding information. It is characterized in that bits of information to be replaced are replaced with corresponding bits. It is also characterized in that the bits to be replaced may be determined depending on the amplitude and phase values of other cosine waves and the amplitude and phase values of the cosine waves. It is necessary that the bits to be replaced and the bits for determining the bits do not overlap.
In the method for extracting information from an image according to the present invention, in order to detect embedded information, a predetermined bit of the amplitude value of a cosine wave constituting the image is read out, and the value or a value corresponding thereto is embedded in the bit of the embedded information. It is characterized by the following. It is also characterized in that the bits to be read may be determined depending on the amplitude and phase values of other cosine waves and the amplitude and phase values of the cosine waves.

[0011]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is an explanatory diagram of a method of embedding information in an image showing one embodiment of the present invention,
FIG. 2 is an explanatory diagram of one embodiment in FIG. In this embodiment, as a representative example of the frequency expression of an image, JPEG
Discrete cosine transform used in compression is taken up, and a digital watermarking method for an image represented by discrete cosine transform is described. "The latest MPEG textbook" ASCII Publishing (1994) 53
As described from page 66 to page 66, in JPEG, an image is compressed for each block composed of 8 × 8 pixels as shown in FIG. As shown in FIG. 1, as a result of compression, 8 × 8
The image 41 of the pixel has a DCT coefficient (Discrete Cosine Tr).
ansformation Coefficient). Therefore, the entire image 43 is represented by a set of 8 × 8 DCT coefficient matrices.

As shown by element 00 in FIG.
The (0,0) component of the T coefficient matrix, that is, the upper leftmost coefficient, represents an average value of 8 × 8 pixel values. The coefficient in the upper left part (element near element 00) represents the amplitude of the low frequency component. This corresponds to the shape of the basic object in the image. The coefficient in the lower right part (element near the element 77) represents the amplitude of the high frequency component. This corresponds to detailed irregularities. The center portion is the amplitude of the medium frequency component, and corresponds to the texture (pattern or texture). To insert information into the JPEG image, the value of the DCT coefficient is changed.
For example, lower several bits are replaced with bits of insertion information for each DCT coefficient. The number of lower bits to be replaced is determined in consideration of the effect on the image. For example, FIG.
As shown in (a), the lower two bits of the DCT coefficients of the three pixels of element 70, element 63, and element 57 are replaced with "10" or "11". Now, the coefficient value of the element 70 is [10
101101], the lower two bits '01' are replaced with '10' or '11'.

Since the target image is a graphic mark, some noise may be added as long as the meaning of the mark is not confused. Therefore, a significant change can be made to the high frequency component. That is, the lower right part (element 77
Is an image corresponding to the detailed unevenness, so that the appearance is hardly changed even after the large replacement. The upper left portion (element near element 00) is an image corresponding to the shape of a basic object, and may be noticeable when a large value is replaced. Substitution of a small value is not so noticeable. When the texture represented by the medium frequency is not regular, the mark visibility can be maintained even if the texture is slightly changed. Therefore, as long as the texture represented by the medium frequency is not regular, the medium frequency component can be significantly changed. Thus, a large amount of information can be embedded by making a large change to the high frequency component and the medium frequency component. Some changes can be made to the low frequency components.

The average luminance of the entire image is calculated by
It corresponds to the average value of the (0,0) component. Therefore, uniformly changing the (0,0) component of all blocks corresponds to shifting the brightness of the image as a whole. this is,
It hardly affects the visibility of the certification mark. Therefore, the element (0,0) component can also be changed. The description so far is based on the upper left block (8 × 8 pixels) in FIG.
Has been described, but by performing the above operation on a plurality of blocks constituting an image, it is possible to change many cosine waves of the image subjected to discrete cosine transform and embed a large amount of information. For example, of the whole image 43,
By replacing the lower two bits of the coefficient values of the elements 00, 07, 70, and 77 of the upper left and upper right and lower left and lower right blocks (8 × 8 pixels) with “10”, Embedding may be performed.

(Embodiment 1) In this embodiment, usually, predetermined bits of the amplitude value of a cosine wave forming an image are replaced with bits of information to be embedded. For example, if the coefficient of the pixel of the element 70 is [10101101], [101011]
10] and the lower 2 bits are replaced with '10'.
As shown in FIG. 2B, if the coefficient of the pixel of the element 70 is

When the value is [00000000], there is a possibility that the pixel content is greatly changed by the replacement. Therefore, when the value is all 0, the target pixel is changed to the next element 71 and [111]
[00101], the lower 2 bits are replaced with '10' and [11]
100110]. When the pixel of the element 70 is the pixel to be replaced, the coefficient of the pixel of the element 60 is [00
000000], it is also possible to change the target pixel from the immediately lower element 70 to the adjacent element 71 and replace the lower two bits of the pixel of the element 71 in the same manner as described above. is there. Further, since the frequency expression by the discrete cosine transform is expressed by the above equation (1), COS
The value of the phase of A greatly depends on the pixel content. Therefore, in the present invention, the bits to be replaced are determined depending on the amplitude and phase values of another cosine wave, or the amplitude and phase values of the cosine wave.

(Embodiment 2) In this embodiment, the bits to be replaced are, for example, the lower two bits, but it is not necessary to directly determine the bits to be replaced. For example, as shown in FIG. 2C, if the bit to be replaced is “01”,
It is also possible to replace with a result of rearrangement, conversion, operation or the like. If the original value '01' becomes '010' due to rearrangement, conversion or operation, this '01'
0 'may be the bit to be replaced. Therefore,
In the present invention, the bits of the amplitude value of the cosine wave forming the image are replaced with bits of information to be embedded or bits corresponding to bits of the information to be embedded (original bits) (bits of the result of rearrangement, conversion or operation). Will be replaced by

(Embodiment 3) In this embodiment, the bits to be replaced are the lower two bits, for example.
As shown in 2), the coefficient value of the element 00 which is the reference pixel is

In the case of [00000000], the change becomes conspicuous if the coefficient value of another element is replaced. Therefore, the replacement of the element is replaced by the lower one bit “1”.
As shown in -1), if the coefficient value of the element 00 is a value other than the above, for example, [00010100], the lower two bits may be replaced with '10'. Therefore, in the present invention, it is possible to make the bit to be replaced different from the bit for determining the bit. As described above, in the present embodiment, when embedding a large number of bits in an image represented by frequency using a digital watermark, replacement is performed with the bits of the information to be embedded. It can be embedded in a short time. Therefore, it is possible to embed multi-bit information such as a digital signature with a digital watermark.

(Information Extraction Method) In the method for extracting information from an image according to the present invention, a predetermined bit of the amplitude value of a cosine wave forming the image is read out and the value or a value corresponding thereto is embedded as information bits. . That is, processing is performed in the reverse order of the embedding method as described in the embodiment, that is, the embedded value is read from the pixel coefficient value of the embedded image block, and if the value matches the embedded value, it is correct. Is determined. The bits to be read are determined depending on the amplitude and phase values of the other cosine waves and the amplitude and phase values of the cosine waves. That is,
Bits may be read using the same criteria as the criteria determined at the time of embedding.

[0019]

As described above, according to the present invention,
Since the bits of the amplitude value of the cosine wave are replaced with the bits to be embedded, much information can be easily and quickly embedded in the image. As a result, it is possible to easily embed multi-bit information such as a digital signature into a small-sized image (graphic mark) by using a digital watermark.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a method of embedding information in an image according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing an embodiment of an information embedding method according to the present invention.

FIG. 3 is an explanatory diagram of an image in a conventional pixel expression.

FIG. 4 is an explanatory diagram of a discrete cosine transform of a frequency expression applied to the present invention.

FIG. 5 is a diagram illustrating a change in amplitude of a cosine wave in a conventional digital watermarking method.

FIG. 6 is an explanatory diagram of a conventional method of embedding a digital signature in a graphic mark.

FIG. 7 is an explanatory diagram of a conventional method for detecting a digital signature from a graphic mark.

[Explanation of symbols]

21 ... Content, 22 ... Digital Signature, 23 ... Graphic Mark, 24 ... Authentication Mark, 25,26 ... Feature Value Extraction, 27 ... Private Key, 28 ... Digital Signature Creation, 29 ...
Digital watermark embedding, 30 synthesis, 27a public key, 2
8a: Digital signature verification, 29a: Digital watermark extraction,
31: judgment, 41: 8 × 8 pixel image, 42: 8 × 8 matrix (DCT coefficient matrix), 43: whole image.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Hisashi Toshima 1-6-27 Shinsuna, Koto-ku, Tokyo Within the Public Works Division, Hitachi, Ltd. (72) Inventor Shinobu Nagakawa 5030 Totsukacho, Totsuka-ku, Yokohama-shi, Kanagawa Address F-term in Hitachi Ltd. Software Division 5B057 AA11 BA24 CA01 CA08 CA12 CA16 CB01 CB08 CB12 CB16 CC02 CE08 CG05 5C076 AA14 BA06 BA09 5J104 AA14 NA15 9A001 EE02 EE03 GG01 HH23

Claims (5)

[Claims] 1. A method for embedding information in an image, wherein, for an image converted into a frequency expression, a predetermined bit of an amplitude value of a cosine wave constituting the image is set to a bit of information to be embedded or to be embedded. A method of embedding information in an image, characterized by replacing bits of information with bits corresponding to the information. 2. The method for embedding information in an image according to claim 1, wherein the bits to be replaced depend on the amplitude and phase values of another cosine wave or the amplitude and phase values of the cosine wave. A method of embedding information in an image, characterized in that it is determined by the following method. 3. The method for embedding information in an image according to claim 2, wherein the bit to be replaced and a bit for determining the bit are different. 4. A method for extracting information embedded by the method for embedding information in an image according to claim 1, wherein a predetermined bit of an amplitude value of a cosine wave forming the image is read and read. A method of extracting information from an image, wherein a value corresponding to the value or a value corresponding to the value is used as a bit of embedded information. 5. The method for extracting information from an image according to claim 4, wherein the bits to be read out depend on the amplitude and phase values of another cosine wave or the amplitude and phase values of the cosine wave. A method for extracting information from an image, characterized in that it is determined.