JP2001218023A - Device and method for embedding forgery detection information and detecting forgery, and recording medium recorded with program conducting the method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device and a method for embedding information and detecting forgery, which can discriminate a picture processing and a forgery behavior and can specify which position in a picture forgery is conducted for every area. SOLUTION: A forgery detection information embedding device 1 divides the picture into plural frequency bands and calculates a conversion coefficient. A random number group is generated by using key data and authentication data is generated. Key data is embedded in the conversion coefficient of MRA and authentication data is embedded in the conversion coefficient of MRR. MRA and MRR to which a embedding processing is conducted, are synthesized and the picture where information is embedded is reconstituted. A forgery detection device 2 extracts key data from MRA where the picture is band-divided and authentication data which is originally embedded is generated. Embedded information is extracted from MRR. The picture is divided into the plural blocks composed of plural pixels, which are previously decided. The group of information embedded in the conversion coefficient of MRR expressing a spatial area similar to the respective blocks is compares/collated with the group of corresponding authentication data. Then, the presence or absence of forgery is judged for every area.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information embedding / falsification detection device and method for falsification detection, and a recording medium on which a program for executing the method is recorded, and more particularly, to a partial falsification of a digital image. In order to detect the presence or absence of tampering and the position of tampering, an apparatus for embedding and extracting (detecting) authentication data for tampering detection in a digital image signal, a method performed by this apparatus, and a program for executing this method are recorded. It relates to a recording medium.

[0002]

2. Description of the Related Art In recent years, the provision of information using the Internet has become popular. In particular, WWW (World
Wide Web) is frequently used as an information transmission / reception service integrating images, sounds, and the like. However, in such an open network environment, an unspecified number of people can easily copy digital information such as images. In addition, these unspecified large numbers of people can easily edit and process digital images using commercially available image processing software. For this reason, even if the digital image is tampered with by a third party during transmission, it is assumed that the tampering is not noticed to the receiver. Therefore, the establishment of falsification detection technology that can determine whether the transmitted digital image has been falsified,
It is urgently required. Conventionally, an electronic authentication technology is known as one of the measures.

FIG. 14 is a diagram illustrating an outline of a conventional procedure of electronic authentication. The transmitting side performs data compression by a hash function on the original digital image to create a digest of the digital image. Next, the sender encrypts the digest using a secret key predetermined by the sender. Then, the transmitting side transmits the original digital image and the encrypted digest to the receiving side through the network. The receiving side, like the transmitting side, first converts the digital image received through the network into
Data digestion by a hash function is performed to create a digest of the digital image. At the same time, the receiving side decrypts the encrypted digest received through the network using the public key determined in advance by the sender to decrypt the digest. Then, the receiving side compares the digest created from the digital image with the decrypted digest, determines that the two digests are the same and determines that they have not been tampered with, and conversely determines that they have been tampered with if they are different. (Electronic authentication).

However, when performing the above-described conventional electronic authentication, the transmitting side needs to transmit two types of data, the original digital image and the encrypted digest, to the receiving side. For this reason, when a large amount of digital images are transmitted, when transmitting two types of data through a network, it is necessary for the transmitting side to appropriately manage which digest corresponds to which digital image. It will be indispensable.

Therefore, as a method capable of performing electronic authentication without performing such management (that is, without transmitting two kinds of data), there is a method using a digital watermarking technique. Digital watermarking is a technique for embedding digital information in digital image data in such a manner that it cannot be perceived by humans. A typical prior art relating to electronic authentication using this digital watermark is disclosed in, for example, the document “P.
ROCEEDINGSOF THE IEEE, VO
L. 87, NO. 7, JULY 1999 p1167
-1180 ".

Here, consider the case where the digital watermarking technique is applied to the above-described digital authentication using a digest. The transmitting side creates a digest of the digital image by a predetermined number of bits from the upper bits of the digital image (each pixel of the digital image). Next, the transmitting side encrypts the digest with a secret key predetermined in the sender. Then, the transmitting side sets the lower bits of the digital image
After embedding the encrypted digest, it sends it to the recipient through the network. The receiving side extracts the encrypted digest embedded in the lower bits of the digital image received over the network. Next, the receiving side performs decryption on the extracted digest using a public key predetermined by the sender, and decrypts the digest. On the other hand, the receiving side, from the upper bits of the received digital image, by a predetermined number of bits,
Create a digest for collation. And the receiving side,
The digest for comparison and the decrypted digest are compared, and if both digests are the same, it is determined that they have not been tampered with, and if they are different, it is determined that they have been tampered with.

[0007]

However, according to the above-mentioned conventional digital watermarking technique, even if the receiving side can discover that the digital image has been tampered with during transmission, it is possible to specify which part of the digital image has been tampered with. There was a problem that it was not possible. In addition, the above-described digital watermark technology embeds specific information by using a high-frequency component which is generally hard to perceive by human eyes. For this reason, when irreversible image processing (compression or decompression) such as JPEG is performed after information is embedded, the embedded information changes and information cannot be read correctly. . That is, there is a problem that it is impossible to distinguish between intentional falsification of a digital image by a malicious person and falsification of a digital image by irreversible image processing that is generally performed without any malicious intention. Furthermore, since a portion corresponding to a high-frequency component of an image is generally an edge or a texture portion, in an image having many flat portions (an image having little change in contrast), information is not embedded in the entire image. Become. Therefore, when the flat portion is tampered with, there is also a problem that embedded information may not be detected.

Therefore, an object of the present invention is to embed specific information not only in a high frequency component of an image but also in the entire image, that is, embed specific information in a transform coefficient of a relatively low frequency component, and to subsequently embed the information. By extracting the information, a falsification detection information embedding device, a falsification detection device, and a falsification detection device capable of distinguishing a falsification act of an image from an irreversible image processing, and capable of specifying a falsified portion. An object of the present invention is to provide a recording medium storing a method and a program for performing the method.

[0009]

A first aspect of the present invention is an apparatus for embedding predetermined information for falsification detection in a digital image signal, and divides the digital image signal into a plurality of frequency bands. A band dividing unit, an authentication data creating unit that creates a pseudo-random number sequence using predetermined key data, and creates authentication data from the pseudo-random number sequence, and a lowest frequency band (MRA) among a plurality of frequency bands. And a key data embedding unit that embeds key data in the transform coefficient, and a frequency band (MR) other than MRA among a plurality of frequency bands.
An authentication data embedding unit for embedding authentication data in the conversion coefficient of R), and a band synthesizing unit for reconstructing a digital image signal in which information is embedded using MRA and MRR after the data embedding process. .

As described above, according to the first aspect, the digital image signal is band-divided into a plurality of layers, and the authentication data is embedded in the conversion coefficient in the MRR. Moreover, the authentication data is created from the pseudo-random number sequence using the key data, and the key data is embedded in the conversion coefficient of the MRA.
As described above, since the information is embedded in the transform coefficient of the relatively low frequency component, even if irreversible image processing is performed, changes in the embedded key data and authentication data are more likely than changes due to intentional tampering. Is also smaller. That is, in the tampering detection device, it is possible to distinguish between irreversible image processing and intentional tampering. Further, in the first invention, it is difficult for a third party who does not know the information such as the embedded frequency band, the conversion coefficient, the reading order of the conversion coefficient, and the key data to decrypt the authentication data. Overwriting and changing files becomes impossible.

A second invention is an invention according to the first invention, wherein a set value T (T is a positive integer) and a set value m
(M is a positive integer equal to or smaller than T), and a value obtained by dividing the transform coefficient by a predetermined quantization step size is preset as q. The authentication data embedding unit sets the absolute value of the transform coefficient to When the absolute value of the conversion coefficient is smaller than the set value T, the conversion coefficient is set to the set value + m or -m according to the bit value of the authentication data to be embedded. When the absolute value is equal to or larger than the set value T, the conversion coefficient is set to an even or odd integer value closest to the value q in accordance with the bit value of the authentication data to be embedded, so that each conversion coefficient of the MRR is authenticated. It is characterized by embedding data.

As described above, the second invention shows a preferable technique for embedding the authentication data in the first invention. By using this method, highly accurate information embedding can be realized with little image quality deterioration.

[0013] A third invention is a device for detecting tampering of a digital image by using tampering detection information embedded in a digital image signal by a specific device, and converts the digital image signal into a plurality of frequency bands. Creating a pseudo-random number sequence using the key data, a band dividing unit to divide, a key data extracting unit that extracts key data embedded in a specific device from MRA transform coefficients among a plurality of frequency bands, An authentication data creation unit for creating authentication data from the pseudo-random number sequence; and embedded information for extracting embedded information embedded in a specific device based on key data from a conversion coefficient of an MRR among a plurality of frequency bands. An alteration determination unit is provided that compares the authentication data with the embedding information and determines whether the digital image has been altered.

A fourth invention is an invention according to the third invention, wherein the falsification determining unit divides the digital image into a plurality of unit blocks composed of a plurality of predetermined pixels. An area-compatible embedded information reading unit that reads out information embedded in the MRR that represents the same spatial area as the unit block from the embedded information for each unit block; A region-corresponding authentication data reading unit for sequentially reading data corresponding to the same position as the embedded information sequentially read by the region-compatible embedded information reading unit from the authentication data; And a block tampering determination unit that determines whether there is tampering with each unit block by comparing the series of embedded information and the series of authentication data for each unit block.

As described above, according to the third and fourth aspects of the present invention, a digital image is divided into unit blocks each composed of a plurality of predetermined pixels, and the same spatial area as each unit block is expressed. The embedding information embedded in the conversion coefficient of the MRR is read, and authentication data that would have been embedded by the falsification detection information embedding device is compared with this embedding information. Thereby, the position of the falsified portion in the digital image can be detected for each area based on the unit block. Also, in the falsification detection information embedding device, since information is embedded in the transform coefficients of relatively low frequency components, even if irreversible image processing is performed, changes in the embedded key data and authentication data may occur. Is smaller than the change due to intentional tampering. That is, irreversible image processing and intentional tampering can be distinguished.

A fifth invention is an invention according to the third and fourth inventions, wherein a value obtained by dividing a set value T and a transform coefficient by a quantization step size and rounding is predetermined as p to extract embedded information. The unit compares the absolute value of the conversion coefficient with the set value T, and determines whether the value of the conversion coefficient is positive or negative when the absolute value of the conversion coefficient is smaller than the set value T. The bit value of the information embedded in the conversion coefficient is extracted based on the conversion coefficient. If the absolute value of the conversion coefficient is equal to or larger than the set value T, it is determined whether the value p is even or odd, and based on the result of the determination, By extracting a bit value of information embedded in the transform coefficient, embedded information embedded therein is extracted from each transform coefficient of the MRR.

As described above, the fifth invention shows a preferred method for extracting the embedded authentication data in the third and fourth inventions. By using this method, highly accurate information extraction can be realized with little image quality deterioration.

A sixth invention is a method of embedding predetermined information for falsification detection in a digital image signal, the method comprising dividing the digital image signal into a plurality of frequency bands, and using predetermined key data. Generating a pseudo-random number sequence and generating authentication data from the pseudo-random number sequence;
A step of embedding key data, a step of embedding authentication data in an MRR conversion coefficient of a plurality of frequency bands, and a step of embedding a digital image signal in which information is embedded using the MRA and the MRR after the data embedding processing. Reconfiguring.

As described above, according to the sixth aspect, the digital image signal is band-divided into a plurality of layers, and the authentication data is embedded in the conversion coefficient in the MRR. Moreover, the authentication data is created from the pseudo-random number sequence using the key data, and the key data is embedded in the conversion coefficient of the MRA.
As described above, since the information is embedded in the transform coefficient of the relatively low frequency component, even if irreversible image processing is performed, changes in the embedded key data and authentication data are more likely than changes due to intentional tampering. Is also smaller. That is, at the time of falsification detection, it is possible to distinguish between irreversible image processing and intentional falsification. Further, in the sixth invention, it is difficult for a third party who does not know information such as the embedded frequency band, the transform coefficient, the read order of the transform coefficient, and the key data to decrypt the authentication data. Overwriting and changing files becomes impossible.

A seventh invention is an invention according to the sixth invention, wherein a set value T and a set value m are predetermined.
The value obtained by dividing the transform coefficient by a predetermined quantization step size is q
The step of embedding the authentication data is a step of comparing the absolute value of the conversion coefficient with the set value T. If the absolute value of the conversion coefficient is less than the set value T as a result of the comparison, the step of embedding the authentication data Setting the conversion coefficient to a set value + m or -m according to the bit value, and, if the absolute value of the conversion coefficient is equal to or greater than the set value T as a result of the comparison, according to the bit value of the authentication data to be embedded, Setting the conversion coefficient to an even or odd integer value closest to the value q, and embedding authentication data in each conversion coefficient of the MRR.

As described above, the seventh invention shows a preferable method for embedding authentication data in the sixth invention. By using this method, highly accurate information embedding can be realized with little image quality deterioration.

An eighth invention is a method for detecting tampering of a digital image by using tampering detection information embedded in a digital image signal by a specific device, wherein the digital image signal is divided into a plurality of frequency bands. Dividing, and from the MRA transform coefficients among the plurality of frequency bands,
Extracting key data embedded in a specific device; generating a pseudo-random number sequence using the key data; generating authentication data from the pseudo-random number sequence; and converting MRR among a plurality of frequency bands. Extracting the embedded information embedded in the specific device based on the key data from the coefficient, comparing the authentication data with the embedded information,
Determining whether the digital image has been tampered with.

A ninth invention is the invention according to the eighth invention, wherein the step of judging the presence or absence of falsification includes dividing the digital image into a plurality of unit blocks each composed of a plurality of predetermined pixels. Step and for each unit block,
MRR representing the same spatial area as the unit block
Sequentially reading out the information embedded in the embedded information, and, for each unit block, data corresponding to the same position as the sequentially read out embedded information in the authentication data. And a step of comparing the sequence of the read-out embedded information and the sequence of the authentication data for each unit block to determine whether there is tampering for each unit block. Features.

As described above, according to the eighth and ninth aspects, the digital image is divided into unit blocks each composed of a plurality of predetermined pixels, and the same spatial area as each unit block is expressed. The embedding information embedded in the conversion coefficient of the MRR is read, and the authentication data that would have been embedded at the time of embedding is compared with the embedding information. Thereby, the position of the falsified portion in the digital image can be detected for each area based on the unit block. Also,
At the time of embedding of the tampering detection information, since information is embedded in the transform coefficient of a relatively low frequency component, even if irreversible image processing is performed, changes in the embedded key data and authentication data are It is smaller than the change due to intentional tampering. That is, irreversible image processing and intentional tampering can be distinguished.

A tenth invention is a invention according to the eighth and ninth inventions, wherein a value obtained by dividing a set value T and a transform coefficient by a quantization step size and rounding is predetermined as p, and embedding information is set. The extracting step includes a step of comparing the absolute value of the conversion coefficient with the set value T. If the absolute value of the conversion coefficient is smaller than the set value T as a result of the comparison, it is determined whether the value of the conversion coefficient is positive or negative. Determining and extracting the bit value of the information embedded in the conversion coefficient based on the result of the determination; and, if the absolute value of the conversion coefficient is greater than or equal to the set value T, the value p is even or odd. Extracting the bit value of the information embedded in the transform coefficient based on the result of the determination, and extracting the embedded information embedded therein from each transform coefficient of the MRR. It is characterized by That.

As described above, the tenth invention shows a preferable method for extracting the embedded authentication data in the eighth and ninth inventions. By using this method, highly accurate information extraction can be realized with little image quality deterioration.

According to an eleventh aspect of the present invention, there is provided a medium in which a method for embedding predetermined falsification detection information in a digital image signal is recorded as a computer-executable program, wherein the digital image signal is stored at a plurality of frequencies. Dividing the band into bands, generating a pseudo-random number sequence using predetermined key data, and generating authentication data from the pseudo-random number sequence; Embedding authentication data in MRR transform coefficients among a plurality of frequency bands, and MRA and M after data embedding processing.
And reconstructing the digital image signal in which the information is embedded using the RR and the RR.

A twelfth invention is an invention according to the eleventh invention, wherein a set value T and a set value m are predetermined, and a value obtained by dividing a transform coefficient by a predetermined quantization step size is defined as q. The step of embedding the authentication data in advance is a step of comparing the absolute value of the conversion coefficient with the set value T, respectively.
If the value is less than the value, the conversion coefficient is set to the set value + m or -m according to the bit value of the authentication data to be embedded. Setting the conversion coefficient to an even or odd integer value closest to the value q in accordance with the bit value of the authentication data, and embedding the authentication data in each conversion coefficient of the MRR.

According to a thirteenth aspect, a method for detecting tampering of a digital image by using tampering detection information embedded in a digital image signal by a specific device is recorded as a program executable by a computer device. A medium, dividing the digital image signal into a plurality of frequency bands; extracting key data embedded in a specific device from transform coefficients of MRA among the plurality of frequency bands; Generating a pseudo-random number sequence using the pseudo-random number sequence and generating authentication data from the pseudo-random number sequence; and embedding information embedded in a specific device based on key data from a conversion coefficient of MRR among a plurality of frequency bands. And the step of comparing the authentication data with the embedding information to determine whether the digital image has been tampered with. Both have recorded a program for execution.

A fourteenth invention is a invention according to the thirteenth invention, wherein the step of judging the presence or absence of falsification includes dividing the digital image into a plurality of unit blocks each composed of a plurality of predetermined pixels. And M for expressing the same spatial area as the unit block for each unit block.
Reading the information embedded in the RR sequentially from the embedded information; and, for each unit block, data corresponding to the same position as the embedded information read serially, Reading out sequentially from inside;
The sequence of the read embedding information and the sequence of the authentication data are
Determining whether there is tampering with each unit block by comparing each unit block.

A fifteenth invention is a invention according to the thirteenth and fourteenth inventions, wherein a value obtained by dividing a set value T and a transform coefficient by a quantization step size and rounding is predetermined as p, and embedding information is set. The extracting step includes a step of comparing the absolute value of the conversion coefficient with the set value T. If the absolute value of the conversion coefficient is smaller than the set value T as a result of the comparison, it is determined whether the value of the conversion coefficient is positive or negative. Determining and extracting the bit value of the information embedded in the conversion coefficient based on the result of the determination; and, if the absolute value of the conversion coefficient is greater than or equal to the set value T, the value p is even or odd. Extracting the bit value of the information embedded in the transform coefficient based on the result of the determination, and extracting the embedded information embedded therein from each transform coefficient of the MRR. Specially To.

As described above, the eleventh to fifteenth aspects of the present invention
A medium on which a program for executing the information embedding / falsification detection method for falsification detection according to the sixth to tenth aspects is recorded. This corresponds to supplying the falsification detection information embedding / falsification detection method according to the sixth to tenth aspects to an existing apparatus in the form of software.

[0033]

FIG. 1 is a block diagram showing a configuration of a falsification detection information embedding device 1 according to one embodiment of the present invention. FIG. 2 is a block diagram showing a configuration of the falsification detection device 2 according to one embodiment of the present invention. In FIG.
The falsification detection information embedding device 1 according to the present embodiment includes a band dividing unit 11, an authentication data creating unit 12, a key data embedding unit 13, an authentication data embedding unit 14, and a band synthesizing unit 15. Prepare. 2, the falsification detection device 2 according to the present embodiment includes a band division unit 11, a key data extraction unit 21,
It includes a key data determination unit 22, an authentication data creation unit 12, an embedding information extraction unit 23, an area corresponding embedded information reading unit 24, an area corresponding authentication data reading unit 25, and a block tampering determination unit 26.

The falsification detection device 2 according to the present embodiment
The band division unit 11 and the authentication data creation unit 12 have the same configuration as the band division unit 11 and the authentication data creation unit 12 of the falsification detection information embedding device 1 according to the present embodiment.
Hereinafter, the same reference numerals are given to the configuration, and the description thereof is omitted. Hereinafter, with further reference to FIGS.
A method for detecting tampering of a digital image performed by the tampering detection information embedding device 1 and the tampering detection device 2 according to the present embodiment will be sequentially described.

First, referring to FIG. 1 and FIGS.
An information embedding method performed by the falsification detection information embedding device 1 according to the present embodiment will be described. FIG. 3 is a flowchart illustrating a process performed by the falsification detection information embedding device 1 according to the present embodiment.

First, with reference to FIGS. 4 to 6, the processing of the band dividing unit 11 that divides a digital image signal into three layers using conventional discrete wavelet transform processing will be described. FIG. 4 is a block diagram illustrating an example of the configuration of the band division unit 11 of FIG. In FIG. 4, the band dividing unit 11
It has first to third band division filters 100, 200 and 300 having the same configuration. The first to third band division filters 100, 200, and 300 divide the input image into four frequency bands, and calculate wavelet coefficients (hereinafter, referred to as transform coefficients) for each frequency band (step S301). Note that a transform coefficient equivalent to band division by discrete wavelet transform can also be obtained by subband division. Band splitting unit 11
Inputs the digital image signal 71 to the first band division filter 100.

The first band division filter 100 converts the digital image signal 71 into four band signals, namely, LL1, LH1, HL1 and HH1 signals, based on the parameters of the horizontal frequency component and the vertical frequency component. (Hereinafter, these are collectively referred to as a first hierarchical signal). The second band division filter 200 receives the LL1 signal of the lowest band among the first hierarchical signals, and furthermore, the LL2 signal, LH2 signal, HL2 signal, and HH2 signal of four bands (hereinafter, collectively referred to as these). Into a second hierarchical signal). Then, the third band division filter 300 outputs the lowest band LL2 of the second hierarchical signal.
Signal, and LL3 signal of four bands, LH3
Signal, an HL3 signal, and an HH3 signal (hereinafter, collectively referred to as a third hierarchical signal).

FIG. 5 shows the first band division filter 1 of FIG.
It is a block diagram which shows an example of a structure of 00. In FIG. 5, the first band division filter 100 includes first to third
It has band division units 101 to 103. The first to third two-band splitters 101 to 103 respectively perform one-dimensional low-pass filters (LPF) 111 to 113, one-dimensional high-pass filters (HPF) 121 to 123, and : Downsamplers 131 to 133 and 1 thinned to 1
41 to 143. First two-band splitting unit 101
Receives the digital image signal 71, performs low-pass and high-pass filtering on the horizontal component by the LPF 111 and the HPF 121, and outputs two signals.
Then, the first two-band splitting unit 101 thins out the low-pass and high-pass filtered signals by 2: 1 using the downsamplers 131 and 141, respectively.
Output to the next stage. The second two-band splitting unit 102 receives the signal from the downsampler 131, performs filtering on the vertical component by the LPF 112 and the HPF 122, respectively, and
Then, after decimating 2: 1 by using 2, the LL1 signal and the LH1 signal are output. On the other hand, the third two-band splitting unit 103 receives the signal from the downsampler 141,
LPF 113 and HPF 123 for vertical component
, And after thinning 2: 1 using the downsamplers 133 and 143, the HL
Two signals, one signal and HH1 signal, are output.

Thus, the first band division filter 10
From 0, the LL1 signal in the low range in both the horizontal and vertical directions,
The LH1 signal in the low band in the horizontal direction and the high band in the vertical direction, the HL1 signal in the high band in the horizontal direction and the low band in the vertical direction, and the HH1 signal in the high band in both the horizontal and vertical directions, that is, four signals, that is, the conversion coefficient is Is output. Note that the second and third band division filters 200 and 300 also perform the same processing on the input signal as described above. As a result of the above-described band division processing by the first to third band division filters 100, 200, and 300, the digital image signal 71 becomes the LL3 signal, the LH3 signal, the HL3 signal, the HH3 signal, the LH2 signal, the HL2 signal, the HH2 signal, The signal is divided into ten frequency bands of the LH1, HL1, and HH1 signals.
FIG. 6 is a diagram expressing these in a two-dimensional frequency domain. Here, the lowest frequency band LL3 signal is called MRA, and frequency bands other than MRA, that is, LH3 signal, HRA
L3 signal, HH3 signal, LH2 signal, HL2 signal, HH
2 signal, LH1 signal, HL1 signal and HH1 signal
Called RR.

In FIG. 6, the vertical axis represents the frequency component in the vertical direction, the lower the lower the frequency, the higher the frequency. The horizontal axis represents the horizontal frequency component, the right the higher the frequency. FIG.
Are data as one image, and the area ratio of the regions matches the ratio of the number of data of each band signal. That is, when the number of data of the LL3 signal, LH3 signal, HL3 signal, and HH3 signal that is the third hierarchical signal is “1”, the number of data of the LH2 signal, HL2 signal, and HH2 signal that is the second hierarchical signal is It becomes “4 (2 × 2 size)”, and the number of data of the LH1 signal, HL1 signal, and HH1 signal as the first hierarchical signal becomes “16 (4 × 4 size)”.

Therefore, for example, regarding one data at the upper left of the LL3 signal, the LH3 signal, the HL3 signal and the H
The upper left one piece of data of the H3 signal is the LH2 signal, the HL2 signal, and the four upper left pieces of the HH2 signal are the LH1 signal, the HL1 signal, and the HH2 signal.
Each of the 16 data in the upper left square of the H1 signal is
The same pixel on the original image will be expressed (in FIG. 6,
This is the part that is painted black.) In other words, 64 pixel data (8 × 8 size) of the upper left square in the image signal expresses the same spatial area as the transform coefficient of each frequency band.

Next, the authentication data creation unit 12 creates a pseudo-random number sequence using predetermined key data, and creates authentication data from the pseudo-random number sequence (step S30).
2). Specifically, the authentication data creation unit 12 determines whether the random real number of the generated pseudo-random number sequence is positive or negative, and if the value is positive, the bit value is “1”; The authentication data is created as "0". For example, the authentication data AD is created from the pseudo random number sequence PN as described below. PN = ｛0.12, -0.23, -1.21,0.23,1.1, -0.34,0.01, -0.5
1, ...,-0.33｝ AD = ｛1,0,0,1,1,0,1,1,
0,..., 0｝ The key data includes, for example, an initial value for generating a pseudo-random number sequence, a type of a random number generation function, information indicating a bit length of the sequence, and the like. In this embodiment, for the sake of simplicity, the key data is a value represented by a data length of 8 bits. Also,
In the following description, it is assumed that the authentication data is a bit stream that is binary-coded into bit values “1” and “0”. It should be noted that the key data information is key information for performing the embedding process, and these information are stored in the falsification detection device 2.
Also used when data is extracted in. Therefore, it is necessary to determine the key data information in advance between both the falsification detection information embedding device 1 and the falsification detection device 2.

Next, the key data embedding unit 13 reads out the MRA conversion coefficients of the image signal divided by the band division unit 11 in a predetermined order, and embeds the key data in the conversion coefficients according to a predetermined method ( Step S303). Various methods can be used to embed the key data. However, if the embedding method proposed by the inventor of the present application (prior application) (Japanese Patent Application Laid-Open No. 11-196262) is used, a high image quality can be obtained with little deterioration in image quality. Accurate embedding can be realized. The embedding method proposed by the inventor of the present application is defined as: q is a value obtained by dividing a transform coefficient by a quantization step size Q.
The value of q is set to the nearest even or odd integer value according to the bit value of the key data corresponding to the conversion coefficient, and the key data is embedded. The predetermined order in which the conversion coefficients of the MRA are read out is key information for performing the embedding process, and these orders are also used when the falsification detection device 2 extracts the key data. Alternatively, the key data may be converted into digital information to which encrypted or error correcting codes have been added, and the digital information may be embedded. Further, when the number of bits of the digital information to be embedded is smaller than the conversion coefficient of the MRA, for example, the digital information may be embedded once and then returned to the first bit of the digital information to be embedded continuously.

Next, the processing performed by the authentication data embedding unit 14 (step S304) will be described with reference to FIG. FIG. 7 is a flowchart illustrating an example of a process performed by the authentication data embedding unit 14 of FIG. First, the authentication data embedding unit 1
4 reads out the conversion coefficients Wi of the LH3 signal shown in FIG. 4 from the MRRs of the image signals divided by the band division unit 11 in a predetermined order (step S701). Next, the authentication data embedding unit 14 sets the absolute value |
It is determined whether or not Wi | is equal to or greater than a predetermined set value T (step S702). If the absolute value | Wi | of the conversion coefficient is smaller than the set value T in the determination in step S702, the authentication data embedding unit 14 determines the bit value of the authentication data corresponding to the conversion coefficient created by the authentication data creation unit 12. , The conversion coefficient Wi is set to a predetermined numerical value + m or −m (step S703). This numerical value m
Can be freely determined as long as it is smaller than the set value T. However, when the numerical value m is reduced, the image quality deterioration is reduced, but the image quality is weakened against attack. When the numerical value m is increased, the image quality deterioration becomes remarkable because the replacement amount is increased, but the image quality deterioration becomes remarkable. What is necessary is just to set suitably according to the level of an image signal, etc. On the other hand, if the absolute value | Wi | of the conversion coefficient is equal to or greater than the set value T in the determination of step S702, the authentication data embedding unit 14 sets the conversion coefficient Wi in the same manner as the key data embedding unit 13 described above. The value divided by the quantization step size Q is set as q, and the value of q is set to the nearest even or odd integer according to the bit value of the authentication data corresponding to the transform coefficient (step S704). By performing the above processing, the bit value of the authentication data is embedded in the conversion coefficient Wi,
A conversion coefficient Wi ′ having undergone the embedding process is created.

When the processing for the LH3 signal is completed, the authentication data embedding unit 14 subsequently reads out the conversion coefficients Wi of the LH2 signal in a predetermined order, and performs the processing of steps S701 to S703 described above. Here, following the processing for the conversion coefficient of the LH3 signal,
Although the processing is performed on the conversion coefficients of the two signals, the order of the processing may be reversed. The embedding of the authentication data is performed by using all the conversion coefficients W in the LH3 signal and the LH2 signal.
It does not need to be performed for i, for example, may be performed vertically and horizontally (see the description of the falsification detection device 2 described later). By doing so, it is possible to reduce image quality deterioration due to embedding of information. The processing order of the LH3 signal and the LH2 signal, and the predetermined order for reading out the conversion coefficients of the LH3 signal and the LH2 signal are key information for performing the embedding process. It is also used when extracting data.
Further, the set values T for comparing the conversion coefficients of the LH2 signal and the LH3 signal do not necessarily have to be the same. Preferably,
The setting value for comparing the conversion coefficient of the LH3 signal is preferably smaller than the setting value for comparing the conversion coefficient of the LH2 signal.
For example, the setting value for comparing the conversion coefficient of the LH3 signal is “7”, and the setting value for comparing the conversion coefficient of the LH2 signal is “1”.
The value m is determined to be “2” or the like.

Next, the processing performed by the band synthesizing unit 15 will be described with reference to FIG.
This will be described with reference to FIG. FIG. 8 is a block diagram showing an example of the configuration of the band synthesizing unit 15 of FIG. 8, the band synthesizing unit 15 includes first to third band synthesizing filters 400, 500, and 60 each having the same configuration.
0 is provided. First to third band synthesis filters 400, 5
00 and 600 receive four frequency band signals,
The signal is synthesized and output as one signal (step S305).
The first band synthesis filter 400 outputs the LL3 signal, LH3
A signal, an HL3 signal, and an HH3 signal are input, and these are combined to create an LL2 signal. Next, the second band synthesis filter 500 outputs the synthesized LL2 signal and LH
Two signals, the HL2 signal and the HH2 signal are input, and these are combined to create the LL1 signal. The third band synthesis filter 600 receives the synthesized LL1 signal, LH1 signal, HL1 signal, and HH1 signal and synthesizes them to reconstruct the digital image signal 72.

FIG. 9 shows the first band combining filter 4 shown in FIG.
It is a block diagram which shows an example of a structure of 00. In FIG. 9, the first band synthesis filter 400 includes first to third two
It has band combining sections 401 to 403. The first to third two-band combining units 401 to 403 respectively
, 413, HPFs 421 to 423, and upsamplers 431 to 4 that insert zero into the signal at a ratio of 2: 1.
33 and 441 to 443, and adders 451 to 453. The first two-band combining unit 401 receives the LL3 signal and the LH3 signal,
1 and 441 are used to convert the signal to a double size signal,
The converted two signals are converted into LPFs 41 with respect to vertical components.
1 and filtered by the HPF 421, added and output. On the other hand, the second two-band combining unit 402
3 and the HH3 signal are input and converted into signals of twice the size using the upsamplers 432 and 442, respectively.
After filtering by PF412 and HPF422, it adds and outputs. Then, the third two-band combining unit 4
03 inputs the outputs of the adders 451 and 452,
2 using upsamplers 433 and 443, respectively.
The signal is converted into a signal of twice the size, the two converted signals are filtered by the LPF 413 and the HPF 423 with respect to the horizontal component, and then added and output.

Thus, the first band synthesis filter 40
From 0, an LL2 signal, which is a second layer signal and is low in both the horizontal and vertical directions, is output. The second and third band synthesizing filters 500 and 600 also perform the same processing on the input signal as described above. The band synthesis unit 15
As described above, the LL3 signal, LH3 signal, HL3 signal, H
H3 signal, LH2 signal, HL2 signal, HH2 signal, LH
The ten frequency band signals of one signal, HL1 signal, and HH1 signal are reconstructed into a digital image signal 72 on which embedding processing has been performed and output.

Next, the falsification detecting device 2 according to this embodiment
Will be described with reference to FIGS. 2 and 10 to 13. FIG. 10 is a flowchart illustrating a process performed by the falsification detection device 2 according to the present embodiment.

In FIG. 2, the band dividing section 11 receives a digital image signal 73 as input. This digital image signal 73
Is the band synthesizing unit 15 of the falsification detection information embedding device 1.
Is a signal that has been subjected to compression encoding / decompression processing, falsification processing, or the like during transmission. The band dividing unit 11 performs a discrete wavelet transform on the input digital image signal 73 to obtain ten frequency bands LL3, LH3, and HL.
3, HH3 signal, LH2 signal, HL2 signal, HH2
Signal, an LH1 signal, an HL1 signal, and an HH1 signal, and the respective transform coefficients are calculated (step S100).
1).

Next, the key data extraction unit 21 executes the MRA conversion coefficient of the digital image signal 73 divided by the band division unit 11 by the key data embedding unit 13 of the falsification detection information embedding device 1. The key data is read out in the same order as extracted, and key data embedded therein is extracted (step S100).
2). Various methods can be used to extract the key data. However, if the technique disclosed in the prior application of the inventor of the present application described above is used, the key data can be extracted with high image quality and little deterioration. it can. That is, the value obtained by dividing the transform coefficient by the quantization step size Q and then rounding the value is p, and it is determined whether the value of p is even or odd, and the bit value of the embedded key data is extracted.

Next, the key data determination unit 22 determines whether or not the key data extracted by the key data extraction unit 21 is the same as the key data used in the falsification detection information embedding device 1. Is determined (step S1003). The determination of the validity is made by comparing the key data extracted by the key data extraction unit 21 with the key data used in the falsification detection information embedding device 1 that is provided in advance. Here, if a plurality of key data are used in the falsification detection information embedding device 1, the key data determination unit 22
May have a plurality of corresponding key data in advance, and compare the key data extracted by the key data extraction unit 21 with the plurality of key data.

In step S1003, if it is determined that the extracted key data is the same as the key data that is previously stored, the key data determination unit 22 performs the processing from step S1004. On the other hand, step S100
In 3, when it is determined that the extracted key data is not the same as the key data stored in advance, the key data determination unit 22 determines that the digital image signal 73 has been tampered with (step S1011). Note that the key data determination unit 22 is not an essential component in the falsification detection device 2. However, by determining the validity of the key data in this way, the reliability of the falsification detection device 2 for detecting falsification of a digital image according to the present invention increases, which is practically preferable.

Next, as described above, the authentication data creating unit 12 uses the key data (initial value for generating the pseudo-random number sequence, random number) used in the falsification detection information embedding device 1 as previously described. A pseudo-random number sequence is created using the generation function type and information indicating the bit length of the sequence, and the authentication data K is created from the pseudo-random number sequence (step S1004).

Next, the processing of the embedded information extraction unit 23 (step S1005) will be described with reference to FIG. FIG.
5 is a flowchart illustrating an example of a process performed by the embedded information extraction unit 23 in FIG. First, the embedded information extraction unit 23
M of the digital image signal 73 divided by the band division unit 11
From the RRs, the conversion coefficients Wi of the LH3 signal and the LH2 signal are converted in the same order as performed by the authentication data embedding unit 14 of the information embedding device 1 for falsification detection, that is, the LH3 signal, L
The signals are read out in the order of the H2 signals and in a predetermined order (step S1101). Next, the embedded information extraction unit 23 determines whether or not the absolute value | Wi | of the read transform coefficient is equal to or greater than a predetermined set value T (step S1102). If the absolute value | Wi | of the conversion coefficient is smaller than the set value T in the determination in step S1102, the embedding information extraction unit 2
3 further determines whether the value of the transform coefficient Wi is positive or negative, and extracts the bit value of the information embedded for each transform coefficient based on the result of the determination (step S1103). On the other hand, when the absolute value | Wi | of the transform coefficient is equal to or greater than the set value T in the determination of step S1102, the embedded information extracting unit 23 quantizes the transform coefficient Wi in the same manner as the key data extracting unit 21. Assuming that the value obtained by dividing by the step size Q and then rounding is p, it is further determined whether the value of p is even or odd, and based on the result of the determination, the bit value of the information embedded for each transform coefficient is extracted. (Step S110
4). By performing the above processing, the embedding information extracting unit 23 embeds the embedding information D (which is embedded in the authentication data that would have been embedded in the falsification detection information embedding device 1), which is information embedded in each transform coefficient. Is extracted (step S10).
05).

Next, the processing of the area corresponding embedded information reading section 24 and the area corresponding authentication data reading section 25 will be described with reference to FIG. FIG. 12 is the same as the unit block at the upper left shaded position in the image when the size of the block for determining whether the digital image signal 73 has been tampered (hereinafter referred to as a unit block) is 32 × 32 pixels. Of the LH3 signal and the LH2 signal expressing the spatial region (corresponding to the thick solid line in FIG. 12). That is, when the unit block size of the digital image signal 73 is 32 × 32 pixels, the corresponding block sizes in the LH3 signal and the LH2 signal are 4 × 4 pixels and 8 × 8 pixels, respectively. It should be noted that the unit block size (32 × 32 pixels) described here is an example, and the unit block size should be used to determine how many block sizes (areas) are used as a unit to modify the digital image signal 73. Can be set arbitrarily.

Now, the predetermined order for reading out the conversion coefficients of the LH3 signal and the LH2 signal (ie, the order in which the information is embedded in the falsification detection information embedding device 1) is
A description will be given by taking an example of a case of every one in the vertical and horizontal directions. In this case, the embedding information corresponding to the unit block of 32 × 32 pixels is 4 in the LH3 signal and 16 in the LH2 signal (in FIG. 12, the portion is blacked out). Therefore,
In the case of this example, the region-based embedded information reading unit 24 extracts the 20 conversion coefficients from the embedded information D extracted by the embedded information extracting unit 23 as the unit for determining whether there is tampering. , Embedded information BD corresponding to a unit block of 32 × 32 pixels (step S1006). Similarly,
The area-corresponding authentication data reading unit 25 includes the authentication data creation unit 1
Out of the authentication data K created in step 2, the 20 authentication data corresponding to the same position as the embedding information BD are set to 32 × 32
The data is read out as authentication data BK corresponding to the unit block of pixels (step S1007).

For example, the embedding information BD and the authentication data BK read by the area corresponding embedded information reading section 24 and the area corresponding authentication data reading section 25 are respectively BD = {1,0,1,1,1,1}. , 0,0,1,1,0,0,1,1,1,1,
BK = {1,1,1,0,1,0,0,1,1,0,0,0,1,0,0,1,0,1,0,1}
0,1,0,1,0,1}. In this example, the four positions of the second, fourth, twelfth, and fourteenth bits are different.

Next, the block falsification judging section 26 first calculates exclusive OR of the bits at the same position between the read embedding information BD and the authentication data BK, and obtains the sum S of the results. (Step S100
8). FIG. 13 is a diagram illustrating an example of an operation based on exclusive OR, and outputs a value “1” when the bit values are different from each other, and outputs a value “0” when the bit values are the same. Then, the block tampering determination unit 26 compares the obtained sum S with a predetermined set value BT, thereby obtaining a partial image divided into unit blocks (32 × 3 in this example).
It is determined whether the tampering has been performed on the two-pixel block (step S1009). For example, the block tampering determination unit 26 determines that tampering has been performed when the obtained sum S is larger than a predetermined set value BT,
On the other hand, if it is smaller, it is determined that the data has not been tampered with. The predetermined set value BT can be freely determined depending on how much falsification is determined, and it is possible to distinguish between image falsification and irreversible image processing by this determination method. In the example of the embedding information BD and the authentication data BK, since four bits are different, the obtained sum S is “4”. Therefore, when the set value BT is set to “3”, since the sum S is larger than the set value BT, it is determined that this unit block has been tampered.

The exclusive OR calculated by the block tampering determination section 26 may be opposite to the above-described logic. That is, if the bit values are different, the value is “0”
, And when the bit values are the same, the value is “1”.
In this case, if the total sum S is smaller than the set value BT, the block tampering determination unit 26 determines that tampering has been performed, and if it is larger, it may determine that tampering has not been performed. Instead of obtaining the sum S of the exclusive OR, the number of bits where the embedded information BD and the bit of the authentication data BK match or the inner product thereof may be obtained. further,
A value “−1” may be given instead of the bit value “0” of the embedding information BD and the authentication data BK, and the inner product thereof may be calculated.

If the result of the determination in step S1009 is that the tampering detection device 2 has tampered with, the position of this unit block (32 × 3
If it is determined that there is no tampering in the unit block (within 2 pixel size), on the other hand, if it is determined that there is no tampering, the information that there is no tampering in the unit block position is stored in the memory or displayed on the display (both are displayed). (Not shown) (steps S1010 and S1011). The position of the falsified portion in the digital image is detected by repeatedly performing the processing of steps S1006 to S1009 for all the unit blocks.

As described above, according to the embodiment of the present invention, in the falsification detection information embedding device 1, the digital image signal is band-divided into three layers, and the LH3
Authentication data is embedded in the conversion coefficients of the signal and the LH2 signal. Moreover, the authentication data is created from the pseudo-random number sequence using the key data, and the key data is embedded in the conversion coefficient of the MRA. Further, in the falsification detection device 2, the digital image is divided into unit blocks each including a plurality of predetermined pixels, and the embedding embedded in the conversion coefficient of the MRR representing the same spatial region as each unit block is performed. The information is read out, and the authentication data that would have been embedded by the falsification detection information embedding device 1 is compared with the embedded information. Thereby, the falsification detection device 2 can detect the position of the falsified portion in the digital image for each area based on the unit block. Further, since the information is embedded in the transformation coefficient of the relatively low-frequency component in the falsification detection information embedding device 1, even if irreversible image processing is performed, changes in the embedded key data and authentication data may occur. Is smaller than the change due to intentional tampering. That is, irreversible image processing and intentional tampering can be distinguished. Further, in the falsification detection information embedding method and the falsification detection method of the present invention, the authentication data can be decrypted by a third party who does not know information such as the embedded frequency band, transform coefficient, read order of the transform coefficient, and key data. The difficulty makes it impossible for a third party to overwrite or replace the embedded information.

Note that the hierarchization by the discrete wavelet transform performed in the band division unit 11 of the above embodiment is not limited to the above-described three hierarchies.
It may be performed any number of times until it becomes one element. The bands used for embedding authentication data are LH3 signal and LH2 signal.
The present invention is not limited to the signal, and may be another band of the MRR or the whole band of the MRR. In this case, similarly, it is necessary to determine the order of the bands to be processed in advance. However, in order to sufficiently exert the useful effects of the present invention, it is preferable to embed the authentication data in the conversion coefficient of only the deeper hierarchical signal. That is, in FIG.
It is most preferable to embed the authentication data in all or a part of the LH3 signal and HL3 signal of the second hierarchical signal and / or the LH2 signal and HL2 signal of the second hierarchical signal.

In the above embodiment, the authentication data embedding unit 14 embeds the authentication data in order for each conversion coefficient read in a predetermined order.
The same authentication data may be repeatedly embedded for each unit block in the transform coefficient in the MRR representing the same spatial area as the unit block.

Further, in the falsification detection information embedding device 1 of the above embodiment, the key data is encrypted using a public key or a common key and then embedded, and in the falsification detection device 2, the encrypted and embedded information is embedded. When decrypting a public key, it is necessary to determine a public key or a common key in advance between both devices.

Note that typically, each function realized by the falsification detection information embedding device 1 and the falsification detection device 2 according to the above-described embodiment is a storage device (ROM, RAM, Hard disk) and a CPU (Central Processing Unit) that executes the program data. In this case, each program data may be introduced via a recording medium such as a CD-ROM or a floppy (registered trademark) disk.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a falsification detection information embedding device 1 according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a falsification detection device 2 according to an embodiment of the present invention.

FIG. 3 is a flowchart showing a process performed by the falsification detection information embedding device 1 of FIG. 1;

FIG. 4 is a block diagram showing an example of a configuration of a band division unit 11 of FIG.

5 is a block diagram illustrating an example of a configuration of a first band division filter 100 in FIG.

FIG. 6 is a diagram expressing each signal obtained by a discrete wavelet transform in a two-dimensional frequency domain.

FIG. 7 is a flowchart illustrating an example of a process performed by an authentication data embedding unit 14 of FIG. 1;

FIG. 8 is a block diagram illustrating an example of a configuration of a band combining unit 15 of FIG. 1;

9 is a block diagram illustrating an example of a configuration of a first band synthesis filter 400 in FIG.

FIG. 10 is a flowchart illustrating a process performed by the falsification detection device 2 of FIG. 2;

FIG. 11 is a flowchart illustrating an example of a process performed by an embedded information extraction unit 23 of FIG. 2;

FIG. 12 is a diagram schematically illustrating conversion coefficients of an LH3 signal and an LH2 signal expressing the same spatial region as a block having a size of 32 × 32 pixels.

FIG. 13 is a diagram illustrating an example of an operation based on exclusive OR.

FIG. 14 is a diagram illustrating an outline of a conventional procedure of electronic authentication.

[Description of Reference Codes] 1 Information tampering detection device 2 Tamper detection device 11 Band splitting unit 12 Authentication data creation unit 13 Key data embedding unit 14 Authentication data embedding unit 15 Band synthesis unit 21 ... key data extraction unit 22 ... key data determination unit 23 ... embedded information extraction unit 24 ... area-corresponding embedded information read unit 25 ... region-corresponding authentication data read unit 26 ... block falsification determination unit 71-73 ... digital image signal 100, 200, 300 ... band division filters 101 to 103 ... two band division units 111 to 113, 411 to 413 ... one-dimensional low-pass filters (LPF) 121 to 123, 421 to 423 ... one-dimensional high-pass filters (HPF) ) 131-133, 141-143 ... downsampler 400, 500, 600 ... band synthesis filter 401-403 ... two-band synthesis unit 431-4 33, 441 to 443: Up samplers 451 to 453: Adders

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) H04N 7/08 H04N 7/08 Z 7/081 7/133 Z 7/30

Claims (15)

[Claims] 1. A device for embedding predetermined information for falsification detection in a digital image signal, comprising: band dividing means for dividing the digital image signal into a plurality of frequency bands; and using predetermined key data. An authentication data generating means for generating a pseudo-random number sequence and generating authentication data from the pseudo-random number sequence; and applying the key data to a conversion coefficient of a lowest frequency band (hereinafter referred to as MRA) among the plurality of frequency bands. Key data embedding means for embedding; authentication data embedding means for embedding the authentication data in a conversion coefficient of a frequency band other than the MRA (hereinafter referred to as MRR) of the plurality of frequency bands; A band synthesizing unit configured to reconstruct a digital image signal in which information has been embedded using the MRA and the MRR; Broadcast implant device. 2. A method in which a set value T (T is a positive integer) and a set value m (m is a positive integer equal to or less than T) are determined in advance, and a value obtained by dividing a transform coefficient by a predetermined quantization step size. The authentication data embedding unit compares the absolute value of the conversion coefficient with the set value T, and when the absolute value of the conversion coefficient is smaller than the set value T, According to the bit value of the authentication data,
The conversion coefficient is set to the set value + m or -m. When the absolute value of the conversion coefficient is equal to or larger than the set value T, the conversion coefficient is set to the value q in accordance with the bit value of the authentication data to be embedded. The information embedding device for falsification detection according to claim 1, wherein the authentication data is embedded in each conversion coefficient of the MRR by setting the integer value to the nearest even or odd integer. 3. An apparatus for detecting falsification of a digital image by using falsification detection information embedded in a digital image signal by a specific apparatus, wherein the digital image signal is divided into a plurality of frequency bands. Dividing means; key data extracting means for extracting key data embedded in the specific device from transform coefficients of the lowest frequency band (hereinafter referred to as MRA) among the plurality of frequency bands; Authentication data generating means for generating a pseudo-random number sequence using the pseudo-random number sequence and generating authentication data from the pseudo-random number sequence; and a conversion coefficient of a frequency band other than the MRA (hereinafter referred to as MRR) among the plurality of frequency bands, Embedded information extracting means for extracting embedded information embedded in the specific device based on the key data, and the authentication data and the embedded information較照 combined with, and a determining alteration determination means whether the alteration of the digital image, tamper detection device. 4. The falsification determining means includes: a block dividing means for dividing a digital image into a plurality of unit blocks each including a plurality of predetermined pixels; and for each of the unit blocks, a same space as the unit block. Area corresponding embedded information reading means for sequentially reading information embedded in the MRR representing a dynamic area from the embedded information; and for each of the unit blocks, the area corresponding embedded information reading means. Area-corresponding authentication data reading means for sequentially reading data corresponding to the same position as the embedded information sequentially read from the authentication data; and By comparing with a series of authentication data for each unit block,
4. The apparatus according to claim 3, further comprising: a block tampering determination unit that determines whether tampering has occurred for each unit block.
Falsification detection device. 5. A value obtained by dividing a set value T (T is a positive integer) and a transform coefficient by a quantization step size and rounding the result is p.
The embedding information extracting means compares the absolute value of the conversion coefficient with the set value T, and when the absolute value of the conversion coefficient is smaller than the set value T, the value of the conversion coefficient Is determined as positive or negative, and a bit value of information embedded in the conversion coefficient is extracted based on the result of the determination. When the absolute value of the conversion coefficient is equal to or greater than the set value T, the value p is By determining whether the number is even or odd, and extracting the bit value of the information embedded in the transform coefficient based on the result of the determination, from each transform coefficient of the MRR, the embedded information embedded therein is obtained. The falsification detection device according to claim 3, wherein the tampering is detected. 6. A method for embedding predetermined information for falsification detection in a digital image signal, the method comprising: dividing the digital image signal into a plurality of frequency bands; and pseudo-random numbers using predetermined key data. Generating a sequence and generating authentication data from the pseudo-random number sequence; embedding the key data in a conversion coefficient of a lowest frequency band (hereinafter referred to as MRA) among the plurality of frequency bands; Embedding the authentication data in a transform coefficient of a frequency band other than the MRA (hereinafter referred to as MRR) among a plurality of frequency bands, and embedding information by using the MRA and the MRR after the data embedding process. Reconstructing a digital image signal on which falsification has been performed. 7. A set value T (T is a positive integer) and a set value m (m is a positive integer less than or equal to T) are determined in advance, and a value obtained by dividing a transform coefficient by a predetermined quantization step size is set. The step of embedding the authentication data is a step of comparing the absolute value of the conversion coefficient with the set value T respectively. As a result of the comparison, the absolute value of the conversion coefficient is smaller than the set value T. Setting the conversion coefficient to the set value + m or -m according to the bit value of the authentication data to be embedded; and, if the absolute value of the conversion coefficient is equal to or larger than the set value T as a result of the comparison, Setting the conversion coefficient to an even or odd integer value closest to the value q in accordance with the bit value of the authentication data to be embedded. And wherein the embedded, alteration detection information embedding method according to claim 6. 8. A method for detecting tampering of a digital image by using tampering detection information embedded in a digital image signal by a specific device, comprising: dividing the digital image signal into a plurality of frequency bands. Extracting key data embedded in the specific device from transform coefficients of a lowest frequency band (hereinafter referred to as MRA) among the plurality of frequency bands; and a pseudo-random number sequence using the key data. And generating authentication data from the pseudo-random number sequence. From the conversion coefficients of frequency bands other than the MRA (hereinafter referred to as MRR) among the plurality of frequency bands, Extracting embedded information based on the authentication data and the embedded information, And a step of determining the presence or absence of falsification, tampering detection method. 9. The step of determining whether or not tampering has occurred: dividing the digital image into a plurality of unit blocks each including a plurality of predetermined pixels; Reading information embedded in the MRR representing the spatial region of the embedded information from the embedded information in sequence, and for each of the unit blocks, the embedded information read in sequence. Reading data corresponding to the same position in sequence from the authentication data, and comparing the read sequence of the embedded information with the sequence of the authentication data for each unit block,
9. The method according to claim 8, further comprising: determining whether there is tampering for each unit block. 10. A method in which a set value T (T is a positive integer) and a value obtained by dividing a transform coefficient by a quantization step size and rounding is predetermined as p, and the step of extracting the embedding information includes: Comparing the value with the set value T. If the absolute value of the conversion coefficient is less than the set value T as a result of the comparison, it is determined whether the value of the conversion coefficient is positive or negative. Extracting the bit value of the information embedded in the transform coefficient based on the result; and, if the result of the comparison indicates that the absolute value of the transform coefficient is greater than or equal to the set value T, determine whether the value p is even or odd. Extracting the bit value of the information embedded in the transform coefficient based on the result of the determination, and extracting the embedded information embedded therein from each transform coefficient of the MRR. Especially To, alteration detection method according to claim 8 or 9. 11. A medium in which a method for embedding predetermined falsification detection information in a digital image signal is a medium recorded as a program executable by a computer device, wherein the digital image signal is divided into a plurality of frequency bands. Creating a pseudo-random number sequence using predetermined key data and creating authentication data from the pseudo-random number sequence; and the lowest frequency band (hereinafter referred to as MRA) of the plurality of frequency bands. Embedding the key data in a transform coefficient of the following; embedding the authentication data in a transform coefficient of a frequency band other than the MRA (hereinafter referred to as MRR) of the plurality of frequency bands; Using the MRA and the MRR to reconstruct a digital image signal in which information is embedded. And up, recording a program for performing at least a recording medium. 12. A set value T (T is a positive integer) and a set value m (m is a positive integer less than or equal to T) are predetermined.
The value obtained by dividing the transform coefficient by a predetermined quantization step size is q
The step of embedding the authentication data is a step of comparing the absolute value of the conversion coefficient with the set value T, respectively. As a result of the comparison, if the absolute value of the conversion coefficient is smaller than the set value T, Setting the conversion coefficient to the set value + m or -m in accordance with the bit value of the authentication data to be embedded; and embedding if the absolute value of the conversion coefficient is equal to or greater than the set value T as a result of the comparison. Setting the conversion coefficient to an even or odd integer value closest to the value q in accordance with the bit value of the authentication data, and embedding the authentication data in each conversion coefficient of the MRR. The recording medium according to claim 11, characterized in that: 13. A medium in which a method for detecting tampering of a digital image by using tampering detection information embedded in a digital image signal by a specific device is a medium recorded as a program executable by a computer device. Dividing the digital image signal into a plurality of frequency bands; and converting key data embedded in the specific device from a conversion coefficient of a lowest frequency band (hereinafter, referred to as MRA) among the plurality of frequency bands. Extracting, generating a pseudo-random number sequence using the key data, and generating authentication data from the pseudo-random number sequence; of the plurality of frequency bands, a frequency band other than the MRA (hereinafter, referred to as MRR) Extracting embedded information embedded by the specific device based on the key data, from the transform coefficients of The authentication data and the compared collating the embedded information, and determining the presence or absence of falsification of the digital image,
A recording medium that records at least a program to be executed. 14. The method according to claim 1, wherein the step of determining whether there is tampering includes dividing the digital image into a plurality of unit blocks each including a plurality of predetermined pixels. Reading information embedded in the MRR representing the spatial region of the embedded information from the embedded information in sequence, and for each of the unit blocks, the embedded information read in sequence. Reading data corresponding to the same position in sequence from the authentication data, and comparing the read sequence of the embedded information with the sequence of the authentication data for each unit block,
14. The recording medium according to claim 13, further comprising a step of determining whether or not tampering has been performed for each unit block. 15. A step of extracting a setting value T (T is a positive integer) and a value obtained by dividing a transform coefficient by a quantization step size and rounding the value, and extracting the embedding information, Comparing the value with the set value T. If the absolute value of the conversion coefficient is less than the set value T as a result of the comparison, it is determined whether the value of the conversion coefficient is positive or negative. Extracting the bit value of the information embedded in the transform coefficient based on the result; and, if the result of the comparison indicates that the absolute value of the transform coefficient is greater than or equal to the set value T, determine whether the value p is even or odd. Extracting the bit value of the information embedded in the transform coefficient based on the result of the determination, and extracting the embedded information embedded therein from each transform coefficient of the MRR. Especially That, the recording medium according to claim 13 or 14.