JP2001028683A - Method for concealing specific information in image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To distribute an image with concealed prescribed information by concealing the information by increasing and decreasing the difference between a pair of prescribed elements in a pixel information string which is obtained by orthogonally transforming a plurality of adjacent pixel pairs at random positions within a prescribed area in the image with respect to a prescribed threshold and then inversely orthogonally transforming the pixel information string to obtain the original image again. SOLUTION: Information is concealed by using a pair of part of objective pixels in a prescribed area in the image. In order to read concealed information, the prescribed order of the pertinent pixel string of an inspected object image is orthogonally transformed and a pair of the prescribed elements of calculated element strings is compared to judge them to be equal or not by a prescribed threshold to obtain a group of concealed bits. Even when only luminance information in pixel information is considered, the possible number of pixel selection combinations for calculating the orthogonal transformation element string in the prescribed area is so huge that it is not decipherable, and it is almost impossible in fact to alter concealed information on the user level.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a method for distributing image information on an information distribution network such as the Internet.
Identification of concealed information such as copyright holder information on the distribution image in the image information, which can be restored when all or part of the concealed information is duplicated and reused to restore the copyright owner information The present invention relates to an information hiding method.

[0002]

2. Description of the Related Art As a method of protecting the copyright of digitized image information on a network and restricting the reuse and resale of a third party by batch duplication or partial duplication (fragmentation), an encryption method has conventionally been used. Are known. This encryption method encrypts the image information to be distributed.In this case, distribution is performed in such a manner that the commercial image is almost eliminated by reducing or distorting the original image by using an index. The receiver is a system in which only a valid licensee can obtain a decryption key and restore information based on the index.

[0003] However, this encrypted distribution system provides a degraded quality index image to a large number of users who do not intend to reuse the received image, which is a serious obstacle from the viewpoint of the entire distribution of image information. Becomes

[0004]

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described disadvantages, and is intended to specify, for example, a copyright holder or a licensee in a distribution image without greatly deteriorating the image quality. It is an object of the present invention to provide a method of concealing specific information in image information, which can conceal and distribute specific information.

[0005]

To achieve the above object, a method for concealing specific information in image information according to the present invention comprises the steps of: using a plurality of pairs of adjacent pixels at random positions in a specific region in an image as a pixel information sequence. Performing an orthogonal transformation, after concealing the concealment information by increasing or decreasing the difference between a pair of specific elements in the pixel information sequence with respect to a predetermined threshold, by orthogonally inverse transforming the pixel information sequence into an image, It is characterized in that specific concealment information is included in an image.

In a normal use state, specific concealment information, for example, copyright holder information is distributed to a user in a form in which the concealment information is concealed. However, if the user reuses this information without having a legitimate license, the copyright owner information is transferred to the reused image information as it is, so the image information considered to be reused at the time of the problem On the other hand, concealment information disclosure processing can be performed to clearly show concealment information such as copyright holder information. As a result, it is difficult to perform unlimited reuse, and it is possible to provide a high-quality image with almost no quality deterioration to the majority of users who do not intend to reuse.

It is known that when image information is divided into pixels, values of color information such as luminance, hue, and saturation between adjacent pixels statistically take close values to each other. Here, using a plurality of adjacent pixel pairs, obtain the information sequence of the orthogonally transformed element pair from the information sequence of the specific order, when looking at the proximity between the corresponding element pair from the obtained information sequence, Again, it takes a proximity value. Therefore, among these elements, for some specific pairs, the difference between them is increased or decreased with respect to a predetermined threshold, that is, deliberately performing equalization or inequality processing, and performing orthogonal inverse transformation. Even when the image information is returned, the deterioration of the image quality is hardly recognized, and as a result, this is often recognized as a weak noise application.

If 1-bit information is allocated to the implementation of equalization / unequalization between the element sequence pairs, it is hidden in the image information. If all the pixels in a specific area of the image are subjected to this orthogonal transformation, even if the pixel order of the orthogonal transformation is not disclosed, only a certain level of resistance to attacks with the intention of destroying concealed information is obtained. I can't. However, if the concealment information is concealed using some pairs of the target pixels, the concealment degree is further improved. Regarding the reading of the concealment information, the orthogonal transformation of the specific sequence of the pixel column of the inspection target image is performed, the calculated specific element pair of the element column is compared, and the determination of equality / unequality is performed with a predetermined threshold value. Thus, a hidden bit group can be obtained.

Here, taking only the luminance information as an example of the pixel information, the number of pixel selection combinations for calculating the orthogonal transformation element sequence of the specific area may be selected to be a huge value that cannot be decoded as described later. Since it is possible, falsifying the hidden information at the user level is practically almost impossible.

Further, it is possible to destroy the concealment information by performing a process of adding random noise to the luminance information of the pixel over the entire screen, but in this case, the quality of the image information itself is deteriorated. It is meaningless because it will have an impact and the value of the product will be diminished and reuse is almost impossible.

Each of the pieces of pixel information constituting the pixel information sequence is selected from a first combination consisting of luminance, hue and saturation, a second combination consisting of Y and color difference, and a third combination consisting of three primary color elements. Thus, at least one element from among the adopted combinations can be used as pixel information.

The number of adjacent pixel pairs including concealment information is smaller than the number of orthogonally transformed pixel pairs, and a position index of a specific region is determined for pixel pairs other than adjacent pixel pairs including concealment information among orthogonally transformed pixel pairs. Can be included.

An adjacent pixel pair has one pixel as a center in a square pixel array, and the other pixel has any one of eight adjacent pixels. In a staggered pixel array, the other pixel has one pixel as a center. A pixel can be any of the six neighboring pixels.

The average value or the weighted average value of the pixel information of a plurality of neighboring pixels is used as the pixel information of the other pixel. The characteristic value of the adjacent pixel is the average value or the weighted average of the characteristic values of the adjacent pixel group. It can be a value.

[0015]

FIG. 1 shows, as an example, a total of 32 pixel blocks (a, b, c, c) of 8 rows and 4 columns in an image.
, E,..., Ma, que, fu, u). One pixel block of the large block M, for example, the pixel block “A” is assumed to be composed of two adjacent pixels α and β as shown in FIG.

Now, a description will be given of a luminance value as an example of a characteristic value of a pixel. Since the pixel pairs are adjacent pixels to each other, the luminance values generally have close values to each other. Of the 32 pixel blocks, 16 pixel blocks corresponding to a half are arbitrarily selected to obtain a sequence relating to the luminance value, and the obtained sequence is subjected to orthogonal transformation. In this case, there may be various combinations of the 16 pixel blocks to be subjected to the orthogonal transformation, but the number of combinations R is mathematically a permutation, and R = 32 × 31 ×... × 17 / ( 16
.Times.15.times..times.2) (1), which is about 600 million. In other words, in order to destroy or alter only the concealment information without deteriorating the image, a specific one must be found from among the huge number of combinations, which is practically impossible. The present invention utilizes that it is close to Large block M in FIG.
The number of combinations can be arbitrarily changed by changing the size of. In the present invention, one specific combination is selected and used from the number of combinations for information hiding.

The orthogonal transform element sequence, for example, the Hadamard transform element sequence of the luminance sequence of the α group obtained in this way is defined as Hα, and the same transform element sequence of the β group is defined as Hβ. Of the 16 column elements selected here, the characteristic values of the n-th element Hαn and Hβn corresponding to each other also indicate close values. Of the differences of these 16 element sequences, for example, the differences of the characteristic values of arbitrary 8 elements are intentionally matched (equalized) or conversely greatly deviated (unequalized) from the threshold value. For example, 1
For example, copyright holder information is concealed as specific information using information of a total of 8 bits for each bit. In that case, the difference of the remaining eight elements can be used as position determination information of a large block described later.

FIG. 3 shows a graph of the above-described eight-element Hadamard transform element sequence. The horizontal axis indicates the element number, and the vertical axis indicates the absolute value of the element characteristic H. Also, ○
The mark indicates the element sequence of the α group, the mark indicates the element sequence of the β group, and T is the threshold value.

Now, an example of concealing 4-bit information “1011” using elements 0 to 3 will be described. The absolute value of the difference between the absolute values of α and β of the element 0 among the Hadamard transform elements of the pixel pair is smaller than the threshold T here.
Assuming that the difference is represented by an absolute value, when the difference> T is associated with “1” and the difference <T is associated with “0”, the element 0 has the hidden information of 1
Therefore, the difference must be larger than T. Therefore, here, an operation is performed where the α element is intentionally reduced and the β element is intentionally increased so that the difference> T. This causes noise to be mixed into the image, but since the neighboring elements take statistically close values, the amount of operation is small and does not cause large noise.

Next, in the element 1, the difference> T. To hide “0”, the α element is made slightly smaller and the β element is made slightly larger, so that the difference <T. Similarly, element 2
The same operation as for element 0 is performed. Since the difference of the element 3 is> T, no special operation is performed. In the same manner, an operation is performed on the elements 4 to 7 in accordance with the concealment information, or no operation is performed. As described above, the large block M in FIG.
Can hide 8-bit information. For example, an arbitrary bit string can be hidden in a specific Hadamard transform element string.

The element information sequence operated so as to include the concealment information in this way is subjected to orthogonal inverse transformation processing, restored to almost the original image information, and used for image distribution to the user.

As shown in FIG. 4, if the large block M is embedded in the target image so as to be widely scattered, even if it is a broken partial image, the large block M is hidden in the large block M. When the copyright owner information exists even at one place in the screen, the copyright owner information can be identified.

In the case of a broken image, that is, a partially duplicated image, the positional relationship between the screen edge of the large block M and the edge of the large block is undefined, but here, 32 blocks (8 pixels × 8 lines)
By performing only the position shift trial, the information of the element pair other than the element pair including the concealment information can be used so that the block position can be specified.

For example, in FIG. 1, among the 16 orthogonal transformation pairs of the sequence “Keluihakonosofininuriuratakuwa”, element numbers 0, 3, 4, 9, 10, 11, 14, and
Equivalent (difference <T) / unequal (difference) of 15 pairs of 8 pairs
T) is used as 8-bit concealment information, and the remaining 1, 2, 2
It is assumed that a specific bit pattern is concealed equally / unequally for the element pairs of 5, 6, 7, 8, 12, and 13.

Here, when an image whose picture edge is unknown, such as the broken image example in FIG. 4, is to be inspected, an arbitrary pixel is regarded as the pixel α of the pixel block “A” in the large block M in FIG. The element numbers 1, 2, 5, 6, among the 16-element orthogonal transformation pairs of the above-mentioned Kerihakonosofininuriuratakuwa
When verifying the equality / unequality of the element pairs of 7, 8, 12, and 13, it does not generally correspond to the specific bit pattern.
Accordingly, by shifting the pixel regarded as α in the above a maximum of 8 pixels in the horizontal direction and a maximum of 8 pixels in the vertical direction,
The correct image edge pixel is found. Therefore, as described above, the remaining pixels can be used for the large block position determination information.

In this case, the probability that the verification bit pattern due to the erroneous position coincides with the correct pattern by chance is 1/256. If this coincidentally coincides, it is further regarded as the pixel α of the pixel block “a”, and the probability that the bit pattern obtained by verifying the adjacent large block coincidence coincidence is also 1
/ (256 × 256), which gradually decreases. By performing verification in a wider range, erroneous determination can be virtually eliminated.

In the block "A" in FIG. 2, the pixel α
And the pixel β are shown adjacent to the image in the horizontal direction. However, the only condition is that the group of pixels β is adjacent to the group of pixels α. For example, as shown in FIG. In this case, it suffices that there is a pixel β corresponding to any one of the eight surrounding pixels for a certain pixel α. Further, as shown in FIG. 6, in a staggered pixel array, it is sufficient that a pixel β corresponding to a certain pixel α corresponds to any one of six surrounding pixels.

By relaxing the condition, the degree of concealment (the degree of difficulty in finding concealed information) can be made larger than the reciprocal of the number of combinations shown in equation (1). FIG. 7 shows an example in which this is applied to an arbitrary 16 blocks in the large block M. Here, the number assigned to each block indicates an example of the information sequence.

In the pixel arrangements of FIGS. 2, 5, and 6, pixels α and β basically show one pixel as a pixel for including concealment information. The average value or the weighted average value of the characteristic values of the pixels may be used. This is because, according to the present invention, the image to which the concealment information has been added is enlarged, reduced, rotated, mirrored, etc. by an image retouching software or the like, and detected by another method not mentioned in the present invention. When corrected, it is unlikely that the corrected pixel and the information content such as luminance will be restored to match the original image. Therefore, it is possible to improve the detection accuracy of the concealment information by expressing the information of the original pixel itself with the average value of the adjacent pixel group in advance.

The configuration information of an image for concealing arbitrary information has been described using luminance as an example.
Each piece of pixel information forming the pixel information sequence includes a first combination including luminance, hue, and saturation, and a second combination including Y and color difference.
And the third combination of three primary color elements, and at least one of the combinations used each time can be used as pixel information. That is, for example, in the first combination, in addition to the luminance alone, the hue or the saturation can be used alone or in combination.

In the three primary color expression system, each color can be used for information hiding means.
Y, BY, R- adopted in V (television)
In the expression method by Y, these can be used as information hiding means.

Various orthogonal transform methods are known in addition to the above-described Hadamard transform. However, since the processing time is relatively short and the practicability is excellent, the discrete cosine transform (DC
T) or discrete sine transform can also be used.

The present invention can be applied not only to a still image but also to a moving image system for continuously transmitting and receiving the still image.

[0034]

According to the present invention, an image can be distributed on a network with almost no deterioration in image quality, and even if an unspecified number of people receive the image, the copyright owner information and the like can be specified. Since the information is transmitted while being concealed and can be reproduced at any time, it is effective in restricting improper duplication / reuse of an image.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram of a large block including a pixel block of 8 rows and 4 columns.

FIG. 2 is an explanatory diagram of a pixel array in a pixel block.

FIG. 3 is an explanatory diagram of a specific information hiding method.

FIG. 4 is an explanatory diagram of a large block in a broken image.

FIG. 5 is an explanatory diagram showing a first positional relationship between a pixel α and a corresponding pixel β.

FIG. 6 is an explanatory diagram showing a second positional relationship between a pixel α and a corresponding pixel β.

FIG. 7 is an explanatory diagram illustrating a relationship between adjacent pixels when various positional relationships between a pixel α and a pixel β are applied to an arbitrary 16 blocks in a large block.

[Explanation of symbols]

 M Large block A ~ P Pixel block α, β Neighboring pixel

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) 9A001 F term (Reference) 5B057 AA11 CA01 CA08 CA12 CA16 CB01 CB08 CB12 CB16 CC03 CD01 CE08 5C059 KK43 LC04 MA22 MA23 MC26 PP14 PP15 PP16 SS08 5C076 AA14 BA06 5D044 AB07 GK08 GK17 5J104 AA14 PA07 9A001 CC07 EE03 HH27 JJ25 KK56 LL03

Claims (5)

[Claims] An orthogonal transformation is performed using a plurality of adjacent pixel pairs at random positions in a specific area in an image as a pixel information sequence,
After concealing the concealment information by increasing / decreasing a difference between a pair of specific elements of the pixel information sequence with respect to a predetermined threshold, the pixel information sequence is returned to an image by performing an orthogonal inverse transform on the pixel information sequence. A specific information concealment information in the image. 2. The method for concealing specific information in an image according to claim 1, wherein each piece of pixel information forming the pixel information sequence is:
The first combination consisting of luminance, hue and saturation, the second combination consisting of Y and color difference, and the third combination consisting of three primary color elements, and at least one of the adopted combinations A method for hiding specific information in an image, wherein one element is used as pixel information. 3. The method for concealing specific information in an image according to claim 1, wherein the number of adjacent pixel pairs including the concealment information is smaller than the number of orthogonally transformed pixel pairs. A method for concealing specific information in an image, wherein a position index of a specific region is included in a pixel pair other than the adjacent pixel pair including the concealment information. 4. The method for concealing specific information in an image according to claim 1, wherein in the square pixel array, one pixel is a center of one pixel and the other pixel is one of eight adjacent pixels. In the staggered pixel array, a specific information concealment method for an image, characterized in that one pixel is centered and the other pixel is one of six adjacent pixels. 5. The method for concealing specific information in an image according to claim 4, wherein an average value or a weighted average value of pixel information of a plurality of pixels around the adjacent pixel is used as the pixel information of the other pixel. A method of hiding specific information in a featured image.