JP2001024876A - Method and device for image processing and storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect an alteration position when the alteration is performed by inputting image position information for instructing an image position at which another data are embedded in image data, generating sub-key information at a random number based on main key information and changing pixel data of the image position of the image data decided based on the image position information based on the sub-key information. SOLUTION: In an extracting device of an electronic watermark, data to be inputted to an input part 21 are embedded image data (x') of the electronic watermark outputted from an electronic watermark embedding device and are inputted to a bit extractor 23. Also, a random number initial value Iv to be inputted to a random number generation part 22 is an initial value to the random number generation part 22, that is, main key information. The random number generator 22 to which the random number initial value Iv (the main key information) is inputted generates a random number progression (sub-key information) Rn and the generated random number progression Rn is inputted to the bit extractor 23. In the bit extractor 23, bit information is extracted from the embedded image data (x') by using this random number progression Rn.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an image processing method and apparatus for generating image data in which a digital watermark is embedded in original image data and / or specifying a falsified portion of the image data, and a storage medium. It is.

[0002]

2. Description of the Related Art In recent years, computers and networks have been remarkably developed, and various types of information such as character data, image data, audio data, and the like have been digitized and handled via computers and networks. Since such data is digital data, it is in an environment where data can be easily copied. Therefore, in order to protect the copyright of such data, a process of embedding copyright information and user information as digital watermarks in digital image data and digital audio data is often performed.

[0003] The digital watermark is a technique for embedding information in digital image data or audio data by performing predetermined processing on the data. By extracting this digital watermark from the data, copyright information and user information can be obtained, and illegal copying can be tracked.

[0004] The method of embedding such a digital watermark can be classified into two methods, a method of embedding in a spatial domain and a method of embedding in a frequency domain. Among them, an example of a method of embedding in a space area is IB as an example of patchwork.
M method (W. Bender. D. Gruhl, N. Morimoto, Techniques fo
r Data Hiding, "Proceedings of the SPIE" .San Jose C
A, USA, February 1995). Examples of the method of embedding in the frequency domain include a method using the discrete cosine transform, the NTT method (“Digital watermarking method in the frequency domain for copyright protection of digital images” by Nakamura, Ogawa, Takashima, etc.), SCIS '97. -26A, 1
In addition to the Discrete Fourier Transform, Defense University's method (Onishi, Oka, Matsui, "Watermark Signature Method for Images Using PN Sequences", SCIS '97 -26)
B, January 1997) and the method of Mitsubishi and Kyushu University using the discrete wavelet transform (Ishizuka, Sakai, Sakurai, "Experimental study on security and reliability of digital watermarking technology using wavelet transform") , SCIS'97-
26D, January 1997) and Matsushita method ("Digital watermark image compression based on wavelet transform, robustness to conversion processing-", Inoue,
Miyazaki, Yamamoto, Katsura, SCIS, 98-3.2. A, 199
Jan. 2008).

[0005] Such a digital watermark needs to be resistant to "attack". Here, “attack” on the digital watermark will be briefly described. This "attack"
Is an attempt to erase or destroy the watermark information, and there are two types. The first one is based on image processing, and includes one that performs image processing such as image compression, enlargement and reduction, cutout, gradation conversion, printout, and scanning on image data in which watermark information is embedded. The second is artificial, and includes adding noise to the image data in which the watermark information is embedded, deleting an image area in which the watermark information is considered to be embedded, and the like. Therefore, the watermark information needs to be resistant to any of such attacks.

[0006] Here, "resistance" means that even if the embedded information is attacked, the embedded information can be accurately extracted.

However, in an application, when an image is attacked, it is not necessary to extract the embedded information. Instead, it is necessary to have a function for specifying the location of the attack. Conceivable. In the following, this function is referred to as the “falsification position detection function”.
I will call it.

As a conventional digital watermarking method having a tampered position detecting function, when an original image is data composed of a plurality of bits per pixel, a least significant bit (LSB) of a plurality of bits is used. The watermark information is embedded by replacing the bits, and when extracting it, the extraction is performed by reading out only the LSB bits. Here, the LSB is bit information located at bit position 0 (minimum order) in information of a plurality of bits forming one pixel, and the bit position is defined as the bit position of the bit in the bit sequence forming the pixel. It indicates the location where it exists. For example, when representing a position where a bit representing 2 ^ 0 = 1 exists, it is called "bit position 0" and 2 ^ 1
When the position where the bit representing = 2 (^ 0 represents the power of 0 and ^ 1 represents the power of 1) is present, it is called "bit position 1". Here, similarly to the case where bit position 0 is referred to as LSB, bit position 7 in 1-byte data is designated as MSB (Most Significant).
tBit). Hereinafter, the expression “bit position” is based on the above definition.

By embedding the watermark information in the LSB, it is possible to specify the falsification position of the attack using the image processing means. This is L
Since the SB represents the minimum size of the bits constituting the pixel, if the tampering is performed, the LSB
Is easy to change, that is, the changed portion can be specified as a falsified portion. If the LSB cannot be correctly extracted, the probability that the LSB is "1" is equal to the probability that the LSB is "0", that is, it is expected that "1" and "0" appear randomly. Therefore, as a result of the extraction, it is possible to verify that a region where “1” and “0” appear randomly is a tampered portion.

[0010] However, the above method is weak against artificial attacks. Since it is fixed and embedded in the LSB, it is conceivable that a third party can easily infer that the digital watermark is embedded by searching for extracting the LSB of all the pixels. Therefore, if a third party is informed that the digital watermark is embedded in the image data, the third party leaves the state of the LSB in which the watermark is rarely embedded and leaves the other bits unchanged. By changing the position data, the image data can be easily falsified. As a specific example, the information of the LSB is stored before the content of the image is falsified,
Thereafter, it is conceivable that the contents of the image are falsified and the position of the falsified image cannot be identified by artificial means such as replacing only the LSB of the falsified image data with information stored in advance.

[0011]

As described above, according to the prior art, when an image in which watermark information is embedded is tampered, the position of the tampered image can be reliably specified. Did not.

The present invention has been made in view of the above conventional example, and has as its object to provide an image processing method and apparatus capable of reliably detecting a tampered position when data is tampered with artificially. And

It is another object of the present invention to specify a pixel position of input original image data, and to rewrite a specific portion of the specified pixel position based on the primary key information so that the input image data has a predetermined value. And an image processing method and apparatus capable of embedding the data.

Another object of the present invention is to provide an image processing method and apparatus capable of detecting falsification processing performed on data in which such predetermined data is embedded and detecting the falsification position. It is in.

[0015]

In order to achieve the above object, an image processing apparatus according to the present invention has the following arrangement.
That is, input means for inputting image position information indicating an image position in which another data is embedded in the image data, random number generating means for generating sub-key information on a random basis based on the main key information,
A pixel changing unit that changes pixel data at an image position of the image data determined based on the image position information based on the sub key information.

Further, the image processing apparatus according to the present invention converts the pixel data at the image position of the image data determined based on the image position information indicating the image position at which another data is embedded in the image data based on the primary key information to a random number. Input means for inputting embedded image data obtained by making a change based on the automatically generated sub-key information, key information generating means for inputting the main key information and generating the sub-key information, Extracting means for extracting the pixel data of the embedded image data based on the sub-key information; and a pixel position based on the pixel data extracted by the extracting means and the sub-key information generated by the key information generating means. Comparing means for comparing the pixel data with the pixel data, and detecting a tampering position of the embedded image data according to a comparison result by the comparing means. That.

In order to achieve the above object, the image processing method of the present invention comprises the following steps. That is, an input step of inputting image position information indicating an image position in which another data is embedded in image data, a random number generating step of generating sub-key information on a random basis based on the main key information, A pixel changing step of changing the pixel data at the image position of the image data determined based on the sub-key information,
It is characterized by having.

Further, according to the image processing method of the present invention, the pixel data at the image position of the image data determined based on the image position information indicating the image position at which another data is embedded in the image data is determined based on the primary key information. An input step of inputting embedded image data obtained by changing based on the sub-key information generated in a random manner, and a key information generating step of inputting the main key information and generating the sub-key information, ,
An extraction step of extracting pixel data of the embedded image data based on the sub key information, and a pixel position based on the pixel data extracted in the extraction step and the sub key information generated in the key information generation step And comparing the pixel data with the pixel data of the embedded image data.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below in detail with reference to the accompanying drawings.

[Embodiment 1] FIG. 1 is a block diagram showing a schematic configuration of an electronic watermark embedding apparatus according to Embodiment 1 of the present invention.

In FIG. 1, data x input to an input unit 11 is multi-valued image data having a predetermined number of bits per pixel. Hereinafter, this data x is referred to as original image data. The data r input to the input unit 12 is 1
Binary image data representing a pixel with 1 bit.
This data is called embedded data. Here, the embedded data may be content that is meaningful to human vision, or content that is meaningless to human vision. If the content is meaningful, the content may be, for example, a logo mark of the author. In this case, the copyright information can be read from the image in which the logo is embedded. Meanwhile,
If the content is meaningless, the ID may be represented by a binary number. Here, the ID means a unique value. For example, in the case of a user ID,
The value represents a value unique to the user, and the device ID represents a value unique to the device.
In this case, it is possible to read user information or device information to be used from an image in which such an ID is embedded.

The data Iv input to the random number generator 13 indicates an initial value to the random number generator 13, that is, primary key information. The embedded data r input to the input unit 12 is input to the embedded image generator 14. The detailed operation of the embedded image generator 14 will be described later. Thus, the embedded image generator 14 outputs image data having an area equal to the original image data x, and this image data is input to the bit replacement unit 15. The original image data x input to the input unit 11 is input to the bit replacement unit 15. The detailed operation of the bit replacement unit 15 will also be described later. In this way, the bit replacement unit 15 outputs a position (for example, the embedded data r ′) specified by the embedded data r in the original image data x.
The image data in which only the specific bit of the pixel information corresponding to “1” is replaced is output, and the image data is output via the output unit 16 as embedded image data x ′.

Next, the operation of each section will be described in detail.

First, the detailed operation of the embedded image generator 14 will be described.

The embedded image generator 14 converts the embedded data r into a size equal to the size of the original image data x based on the input embedded data r. Hereinafter, the image data output from the embedded image generator 14 will be referred to as embedded image data r ′. Various methods can be considered for the size conversion in the embedded image generator 14, such as a method of repeatedly arranging the embedded data r or a method of enlarging (magnifying) the embedded data r to the size of the original image data x. . The embedded image data r 'output from the embedded image generator 14 must also be binary data having one bit of data per pixel.

Next, the detailed operation of the random number generator 13 will be described.

The random number generator 13 receives a random number initial value Iv, ie, primary key information. When the random number initial value Iv is used as public information in order to safely use the processing according to the first embodiment, the random number generator 13 used in the first embodiment
The method of generating the random numbers must be private or use other private information. On the other hand, if the random number generation method in the random number generator 14 is public, the random number initial value Iv must be used as non-public information. That is, the random number sequence input to the bit replacement unit 15, that is, the sub-key information Rn needs to be secret information. Which of these is made public information and which is made non-public information can be determined according to each application.

Next, the detailed operation of the bit replacement unit 15 will be described.

The bit replacing unit 15 is a part for performing a so-called digital watermark embedding process for embedding the embedded image data r 'in the original image data x. In this bit replacement unit 15, the embedded image data r ′ output from the embedded image generator 14 and the corresponding pixel position of the original image data x output from the input unit 11 (for example, the corresponding bit of the embedded data r ′) Is “1”),
A certain bit of the plurality of bits (8 bits in this embodiment) that is the pixel value is replaced. This bit replacement process is performed on an area in which the luminance value data of the original image data x is developed on a bit plane. Now, for example, when the original image data has 256 gradations (8 bits), one of these 8 bits is selected, and this bit is replaced by the corresponding bit of the embedded image data r ′. The bit position to be embedded (selected) is determined by the sub-key information Rn output from the random number generator 13.

For example, if the original image data x is 8 per pixel,
When represented by a binary code of bits, the random number generator 13 outputs a random number data string Rn having a value of “0” to “7”. The bit position corresponding to the value of the random number sequence Rn indicates the bit position to be subjected to the above-described bit replacement.

Here, care must be taken in selecting the bit position to be replaced. For example, when the selected bit position has a value close to the MSB, the pixel value of the embedded image data x ′ is significantly different from the corresponding pixel value of the original image data x, and thus the embedded image data x Is expected to deteriorate compared to the original image data x. This is, for example, in the case of image data composed of 8 bits per pixel, MSB
Is caused by a change of 2 7 = 128 in the luminance value. Further, if only the bit position limited to the LSB is selected, as described in the above-described related art, the resistance to the artificial attack is weakened. That is, for example, the bit replacement target is LS
In the case where B is limited to only the bit position 0, the replaced bit is considered to be easily rewritten. Alternatively, if the target positions for bit replacement are bit position 0 and bit position 1, rewriting the bit position requires a square search of (XS × YS). Here, XS and YS indicate the horizontal size and the vertical size of the original image data x.

In general, if the bit replacement target position is from bit position 0 to bit position n, rewriting the bit position requires a search of (XS × YS) to the nth power. Therefore, by the search described above, the embedded information (that is, the bit position in all the embedded pixels) is obtained.
Is found, after tampering is performed to change the content of the image, the embedded information is replaced with the original embedded information, so that the bit extractor 23 described later It is possible to disguise that there is no image. From this point, it can be said that the bit position to be selected as a bit replacement target is a bit position close to the LSB. By selecting a bit close to the LSB and embedding the embedding data r ′ in this way, visual quality deterioration from the original image data x is suppressed, and
In addition, when tampering has been performed by image processing, it is possible to verify the tampering and further specify the tampered position. That is, it is possible to prevent the fact that the tampering cannot be verified or the tampering position cannot be specified despite the tampering of the image.

By performing the digital watermark embedding process by the process described above, the generated embedded image data x ′ is output via the output unit 16.

FIG. 2 is a block diagram showing a schematic configuration of the digital watermark extracting apparatus according to Embodiment 1 of the present invention.

In FIG. 2, the data x 'input to the input unit 21 is the digital watermark embedded image data x' output from the digital watermark embedding device in FIG. The data I input to the random number generator 22
v is an initial value to the random number generator 22, that is, primary key information. The value of the primary key information Iv is determined by the random number generator 13 in FIG.
Must be exactly equal to the data Iv input to the

The embedded image data x ′ input to the input unit 21 is input to the bit extractor 23. The random number generator 22 to which the random number initial value Iv (primary key information) has been input generates a random number sequence (sub key information) Rn, and the generated random number sequence Rn is input to the bit extractor 23. Bit extractor 2
In step 3, bit information is extracted from the embedded image data x 'using the random number sequence Rn. Here, the random number sequence Rn output from the random number generator 22 has the same configuration as the random number generator 13 in FIG.
The bit position targeted by 3 also matches the bit position targeted by bit replacer 15 in FIG. Therefore, the bit information extracted by the above processing, that is, the image data x ″ indicating the bit extraction result is equal to the original image data x in FIG. 1 when the embedded image data x ′ is not falsified. Should be.

On the other hand, when the embedded image data x 'is falsified, the embedded data r is erroneously detected with respect to the falsified position. Here, whether the image corresponding to the bit extraction result detected from the embedded image data x 'is correct or incorrect depends on what kind of embedded data r is embedded in FIG. The way is different.

For example, if the embedded data r is embedded with contents that are meaningful to human vision, even if the tampering position detector 24 in FIG. 2 is not used (path A), Can be determined. If erroneously detected, it is expected that the binary codes "1" and "0" will appear with equal probability, depending on the type of tampering performed. That is, the altered portion should be seen by human eyes as an image having a property close to white noise.

Further, in FIG. 2, the embedded data r is
If the data can be input to the bit extractor 23, the embedded data r is input to the tampered position detector 24, so that the above determination can be made by the tampered position detector 24. This is because in FIG.
4, that is, the route B.

Here, the tampered position detector 24 receives the embedded data r and the image data (B) indicating the bit extraction result output from FIG. Embedded data r
Is input to the tampered position detector 24 after a predetermined process is performed by the embedded image generator 26. The processing performed by the image generator 26 must be similar to the processing performed by the embedded image generator 14 in FIG. Thus, the tampered position detector 24 compares the bits at the same coordinate position of the two input image data,
"1" is output when they are equal, and "0" is output when they are not equal. As a result, “0” is output as a bit extraction result image at a tampered position, and “1” is output at a position not tampered. The image data c indicating the bit extraction result can be used by an application to specify a tampering position, and further, it is possible to specify a tampering position by human eyes.

In the case where the tampering position is specified by human eyes, this method can detect minute tampering more accurately than the above-mentioned method.

Further, when the embedded data r is a single color of white, the tampered position can be expressed more easily by making the pixels for which tampering has been detected black pixels.

[Second Embodiment] Next, a second embodiment of the present invention will be described. In the second embodiment, the first embodiment is applied to compressed image data. In the first embodiment, the input image data to be embedded is multi-valued image data having a predetermined number of bits per pixel (8 bits in the above example). In contrast, in the second embodiment, the input image data to be embedded is encoded image data. Note that the encoded image data includes JPEG
Also includes image data compressed by. Hereinafter, a case will be described with the second embodiment where the encoded data is JPEG data.

First, JPEG encoding will be described.

FIG. 3A is a block diagram illustrating an example of a JPEG encoding procedure.

First, the image data of the still image to be encoded is 8
Divided into × 8 pixel blocks, DCT for each block
(Discrete cosine transform) is performed (image converter 301).
Hereinafter, a block subjected to the DCT (Discreate Cosine Transform) is referred to as a DCT coefficient block, one coefficient of the DCT coefficient block is referred to as a DCT coefficient, and a set of DCT coefficient blocks of one image is referred to as a DCT coefficient block group.

Next, the DCT coefficient block group is quantized by the quantizer 302 using an arbitrary quantization table. Hereinafter, a block obtained by quantizing this DCT coefficient block is referred to as a quantized DCT coefficient block,
A set of CT coefficient blocks is called a quantized DCT coefficient block group.

Then, the quantized DCT coefficient block group is Huffman-coded (entropy coder 303). The Huffman-encoded data becomes JPEG data.
The Huffman table used at this time may be a table prepared in advance or a table created for each image.

The JPEG data includes the Huffman-encoded data, the quantization table and the Huffman table used at the time of compression (see FIG. 4 described later). Where the quantization D
The CT coefficient block is obtained by dividing each DCT coefficient block by a value determined by a quantization table. For example, a quantization table for a luminance component or a quantization table for a chrominance component is shown in FIG.
(B).

These quantization tables are configured in such a manner that more bits are allocated to the low-frequency component and fewer bits are allocated to the high-frequency component in consideration of human visual characteristics. Therefore, when the value of the DCT coefficient block is “−36”, the quantization table value is “18”.
If so, the quantized DCT coefficient block value is "-2", and if "99", the quantized DCT coefficient block value is "0".

The Huffman coding is performed separately by dividing the quantized DCT coefficients into DC coefficients and AC coefficients. DC
The coefficients are encoded by using a DC table to calculate the difference between the coefficient and the DC coefficient of the immediately preceding block group.

The AC coefficients are rearranged in a zigzag scan order as shown in FIG. 6 to determine the continuous length (run length) of the coefficient “0”,
Encoding is performed using a table in which coefficient values other than “0” are combined.

Next, an example of the JPEG data is shown in FIG.

FIG. 4 is a diagram showing a configuration example of data compressed by the JPEG sequential method.

The sequential system is a system in which clear images appear in order from the top when decoding is performed. On the other hand, the system in which the entire image is displayed unclear at first, and the system which gradually becomes clear is It is called a progressive system. The JPEG data is formatted by a unique 2-byte code in data called a marker.

The first “SOI” marker is a JPEG
Indicates the start of data. The next “DQT” marker portion indicates the definition of the quantization table, and stores the quantization table used at the time of compression after the “DQT” marker. The next “SOFO” marker section indicates “SOF” when compressed by the sequential method using DCT.
Using the "O" marker, the size of the compressed image, sampling rate, number of components,
A parameter at the time of compression such as an identifier of a quantization table for each component is stored. The next “DHT” marker section indicates the definition of the Huffman table.
The Huffman table used at the time of compression is stored after the “HT” marker. The next “SOS” marker part is
The Huffman code which actually encoded the image is stored.
It should be noted that the Huffman code here is differently encoded in the DC component and the AC component of the DCT coefficient.

In JPEG, the DC component of the DCT coefficient expresses the difference value between the DC components of adjacent DCT blocks in Huffman code. On the other hand, for the AC component, the combination of the zero run length and the effective coefficient of the sequence zigzag-scanned in the block is represented by a Huffman code without using the difference value. This Huffman code,
A combination of codes expressing the magnitude of the effective coefficient with the minimum number of bits is encoded data. In the following description, the code corresponding to the former is “GC”, and the code corresponding to the latter is “VC”.
Is defined. Here, the effective coefficient is a coefficient that is not zero.

The Huffman coding of the AC component will be described with an example.

For example, consider a case in which a Huffman coding is performed on a coefficient having a value “−3” located immediately to the right of a DC component as an AC component. In this case, the coefficients are grouped using a table as shown in FIG. That is, for example, when the value is “−3”, “2” is selected as the group number. Further, the zero run length is "0". Huffman coding is performed using the Huffman table shown in FIG. 8 based on the combination (0, 2) of these group numbers and zero run length.

In the case of this example, the Huffman code "01" is selected and stored as "GC". After this, "V
"0" which represents "-3" with the minimum number of bits as "C"
"0" is subsequently stored. As a result, the Huffman code “0100” is selected for the AC component “−3” in this example.

After the "SOS" marker, a header containing information such as an identifier in the Huffman table for each component continues for several bytes, followed by a Huffman code for encoding the image. Finally, the end of the encoded image data is indicated by the “EOI” marker.

In this example, the order of the markers is shown in FIG. 4, but it is not necessary for the actual data to be in this order.
In some cases, the number of markers is two or more.

FIG. 3B is a block diagram showing a configuration example of a decoding device for JPEG data compressed in the procedure of FIG. 3A.

In this decoding process, first, Huffman decoding of Huffman code is performed by using a Huffman table in JPEG data, and decoded into a group of quantized DCT coefficient blocks (entropy decoder 304). Next, inverse quantizer 3
05, the quantized DCT coefficient block group is
Inverse quantization using a quantization table of G data, DCT
Decode into coefficient block group. And then the image inverse converter 3
06, the block group is assigned to IDCT (Inverse DC
T: inverse DCT) to return to a block of 8 × 8 pixels and reconstruct it. By such a procedure, JPE
Decoded image data can be obtained from the G data.

FIG. 9 is a block diagram showing a schematic configuration of a digital watermark embedding device according to Embodiment 2 of the present invention.

In FIG. 9, input data (Code) input via the input unit 901 is JPEG encoded data having a data format as shown in FIG. The random number initial value Iv and the random number generator 905 are the same as the random number generator 13 of the first embodiment. The input JPEG encoded data (Code) is analyzed by the analyzer 902 according to the JPEG encoding format. Here, the analysis means that the input bit sequence is a JPE
According to the G encoding format, it is understood as meaningful data as encoded data. Among these, the Huffman code of the AC component to be embedded with the digital watermark is input to the replacing unit 903 for each DCT block. The replacement unit 903 replaces a Huffman code according to the bit information of the embedded data r for each DCT block. Here, the code to be replaced with the Huffman code preferably has an AC component in the block as much as possible.

This means that, in the scan data input to the replacement unit 902, a code located as far back as possible is targeted. Because the scan data is DC
This is because the zigzag scan is performed in the T block, and the data located later in the scan data has a higher frequency component.

Next, a method of performing a replacement process on a selected code will be described.

First, consider the case where "GC" is to be replaced (method 1).

As described above, this is expressed by a combination of the zero run length and the effective coefficient.
Therefore, if the value of the zero run length changes as a result of the replacement of the Huffman code, correct decoding cannot be performed at the time of decoding. This is because the position of the effective coefficient cannot be determined correctly. As one method to solve this, in JPEG encoding,
In the case where the Huffman table as shown in FIG. 8 is used, it is conceivable to limit the replacement to the horizontal direction of this table. For example, replacing the code “00” with the code “01” can be considered. Further, after the replacement in the horizontal direction is performed, the subsequent “VC” needs to be changed. This is because replacing the code “00” with the code “01” does not change the zero run length, but changes the group number. This can be solved by selecting a code that represents a value that minimizes the difference between the value represented by the code before replacement and the value represented by the code from the group after replacement.

Regarding the horizontal replacement in the present embodiment, for example, when the bit information of the embedded data r is “0”, the bit number is shifted to the nearest odd group number.
On the other hand, when the bit of the embedded data r is “1”, it is realistic to replace the bit number with the nearest even group number from the viewpoint of minimizing image quality deterioration.

Next, consider the case where "VC" is to be replaced (method 2).

As described above, the effective coefficient is represented by the minimum number of bits. Therefore, there is no relation regarding the zero run length as in the case of “VC”. Regarding this code replacement, any method can be selected as long as the method does not change the number of bits. This is because if the number of bits does not change, the group number does not change, that is, decoding can be performed correctly. Regarding the actual replacement, it is conceivable to replace the bit at the bit position with the sign with the bit of the embedded original image. The bit position at this time can be selected by a random number.

As described above, bit replacement is performed by the method described above, and JPEG data is output via the output unit 904.

FIG. 10 is a block diagram showing a schematic configuration of a digital watermark extracting apparatus according to the second embodiment.

In FIG. 10, input data x input via input unit 1001 is converted into a JPEG image in which a digital watermark is embedded by the digital watermark embedding apparatus shown in FIG.
This is encoded data x. The random number initial value Iv and the random number generator 905 are the same as the random number generator 13 described in the first embodiment. Here, the input JPEG encoded data x is analyzed in the analyzer 1002 according to the JPEG encoding format. The Huffman code of the AC component in which the digital watermark is embedded is extracted by the extractor 1003 for each DCT block.
Is input to

The extractor 1003 extracts a digital watermark from the encoded data x for each DCT block. This extractor 1003 includes a random number sequence R obtained from a random number generator 905.
With n, which encoded coefficient is embedded in the block with the digital watermark is specified. Next, a digital watermark is extracted from the code representing the specified coefficient.

Consider a case where both “VC” and “GC” are to be replaced (corresponding to method 1).

This corresponds to “VC” and “G”, as described above.
The digital watermark is embedded in both of “C” and “C”, and the bit can be determined based on the odd / even group number to which the code belongs.

In the input JPEG data, “HD
Using the Huffman table stored after the "T" marker, a group number is obtained from the selected Huffman code, and if this is an even number, a bit "1" is output; if it is an odd number, a bit "0" is output. Output.

Next, consider the case where "VC" is to be replaced (corresponding to "method 2"). This can be extracted by specifying the embedded bit position in the selected code using the random number Rn, and outputting the bit information of the specified bit position. Through the above processing, the digital watermark is extracted and output as an image indicating the bit extraction result via the output unit 1004.

With respect to a method of verifying whether or not tampering has been performed from the bit extraction result image, a method similar to that described in the first embodiment can be used.

Third Embodiment Next, a third embodiment of the present invention will be described. This Embodiment 3 is Embodiment 1
Alternatively, this is a method of realizing the second embodiment by a combination with a hash value.

Generally, it is possible to verify whether digital data has been falsified by using the hash value.

Next, the hash value will be described.

The hash value h is a hash function f: x →
h is a short output h which is the compression value of the long input example x determined by h. Further, the hash function f is a one-way function, and different inputs x, x ′ satisfying r (x ′) = r (x).
Is difficult to find. A typical example of the hash function f is MD5 (Message Digest).
5) and SHA (Secure Hash Algorithm). The details of this hash function are described in "Introduction to Cryptography" by Eiji Okamoto (Kyoritsu Shuppan Co., Ltd.).

As described above, falsification verification of digital data can be performed using hash values having the above-described properties.

When the hash value h of a certain digital data is known, the hash value h of the digital data is calculated again in order to verify whether the data has been falsified. This can be realized by comparing with the hash value h. That is,
If they have not been tampered with, they match, and if they have been tampered, they do not match.

With the method described above, it is possible to verify whether digital data has been falsified. However, in a case where image data is used as digital data used in this method, in a certain application, if the image data has been tampered with, it may be necessary to know which part of the image data has been tampered with.
In such a case, the above method is not sufficient. This is because, in the above-described method of calculating and comparing the hash value h, when at least one bit of the image data is falsified, a comparison result indicating that the hash values h do not match can be obtained. On the other hand, there is no means for obtaining any information as to “where 1 bit has been tampered with”.

Therefore, in the third embodiment, a method is proposed in which tampering detection based on the hash value h is combined with tampering detection according to the above-described first or second embodiment.

FIG. 11 is a block diagram showing a configuration of a digital watermark embedding device according to Embodiment 3 of the present invention.

The input to the present apparatus is multi-valued image data or coded data x having a predetermined number of bits per pixel, embedded data r, and random number initial value k as original data. Here, when the input data x is multi-valued image data having a predetermined number of bits per pixel, the digital watermark embedding device 1101 of FIG. 11 uses an apparatus of the system described in the first embodiment. On the other hand, when the input data x is encoded data, the digital watermark embedding device 1101 in FIG. 11 should use the device of the system described in the second embodiment.

The embedded data r and the random number initial value k
May be the same as those described in the first and second embodiments.

With the configuration as described above, the embedded data r is embedded in the input data x by the digital watermark embedding unit 1101. Here, the data output from the digital watermark embedding device 1101 is referred to as embedded data x1. The embedded data x1 is input to the arithmetic unit 1102 and the synthesizer 1103. The arithmetic unit 1102 performs the above-described hash operation or the like on the embedded data x1, and outputs a hash value h or the like as the operation result. Also, the hash value h of the embedded data x1 output from the digital watermark embedding unit 1101 and the embedded data x1 output from the arithmetic unit 1102 is input to the synthesizer 1103, and these are synthesized. In this synthesis, the hash value h
May be written to a location where additional information corresponding to the format of the input data x can be described. For example,
If the input data x is JPEG encoded data,
The embedded data x1 is also JPEG encoded data. In this case, since COM is defined as a marker for a comment in the JPEG code,
A hash value h can be described after M_COM.

Similarly, when the input data x is multi-valued image data having a predetermined number of bits per pixel,
It is conceivable to write additional information in a place where additional information can be described in the image format used by it. For example, if the input data x is in the PBM format,
Since the data up to the line feed code following the symbol “#” is regarded as a comment, the hash value h is stored in this area.
Can be described.

Further, if the input data x is a FlashPix file
In the case of the format, the hash value h can be stored in the property set as attribute information.

As described above, the combiner 1103 outputs the data x 'in which the hash value h and the embedded data x1 are combined.

Hereinafter, the output from the synthesizer 1103 will be referred to as embedded data x 'including the hash value h.

FIG. 12 is a block diagram showing a configuration of a tampering detection device according to the third embodiment.

The inputs to this device are the embedded data x 'including the hash value h and the random number initial value k output from the digital watermark embedding device shown in FIG. If the random number initial value k is equal to the value input in the device shown in FIG. 11, it is possible to normally extract a digital watermark. The embedded data x ′ including the input hash value h is input to the analyzer 1201. The analyzer 1201 performs an analysis according to the input image format, and outputs a hash value h and embedded data x1. Among them, the embedded data x1 is input to the arithmetic unit 1202 and the switching unit 1204. The arithmetic unit 1202 performs the same operation as the arithmetic unit 1202 in FIG. 11, and outputs the hash value h ′ of the embedded data x1 as its output. In this way, the hash value h ′ of the embedded data x1 output from the arithmetic unit 1202 and the hash value h output from the analyzer 1201 are input to the comparator 1203. Then, the comparator 1203 compares whether or not the two input hash values are equal, and outputs a comparison result y. For example, when the two hash values are equal, the control signal y = "0" is output, and when the two hash values are not equal, the control signal y = "1" is output. By this control signal y, FIG.
It is possible to verify whether the data x ′ input to the device is falsified.

Further, the control signal y output from the comparator 1203 may be input to the switch 1204. The switch 1204 inputs the embedded data x1 output from the analyzer 1201 to the digital watermark extractor 1205, and determines whether to continue the processing. That is, when the control signal y output from the comparator 1203 is “1”, the switch 1204 inputs the embedded data x1 to the digital watermark extractor 1205, and the control signal y is “0”. In a certain case (when they match), an operation is performed to terminate the processing without inputting the embedded data x1 to the digital watermark extractor 1205.

As a result, if the input data has been tampered with, the digital watermark is extracted, and if the input data has not been tampered, the processing can be terminated.

When the control signal y output from the switch 1204 is "1" (when they do not match),
If the embedded image has been tampered with, the embedded image is input to the digital watermark extractor 1205. From the embedded image input to the electronic watermark extractor 1205, an embedded electronic watermark is output using a random number initial value k. The operation of the digital watermark extractor 1205 is the same as the operation described in the first and second embodiments, and a bit extraction result image capable of specifying a tampering position is output. By performing the above-described processing, it is possible to verify whether or not the input data has been tampered with. If the data has been tampered with, it is possible to specify which part of the image has been tampered with.

The present invention is not limited to an apparatus and a method for realizing the above-described embodiment and a method for performing a combination of the methods described in the embodiment, and a computer (CPU or CPU) in the above-described system or apparatus. MP
U), a program code of software for realizing the above-described embodiment is supplied, and the computer of the system or the apparatus operates the various devices according to the program code to realize the above-described embodiment. It is included in the category of the present invention.

In this case, the software program code itself implements the functions of the above-described embodiment, and the program code itself and means for supplying the program code to the computer, specifically, A storage medium storing the above program code is included in the scope of the present invention.

As a storage medium for storing such a program code, for example, a floppy disk, hard disk, optical disk, magneto-optical disk, CD-ROM, magnetic tape, nonvolatile memory card, ROM, etc. can be used.

In addition to the case where the computer controls various devices in accordance with only the supplied program code to realize the functions of the above-described embodiment, the program code operates on the computer. Such a program code is included in the scope of the present invention even when the above-described embodiment is realized in cooperation with an OS (Operating System) or other application software.

Further, after the supplied program code is stored in a memory provided in a function expansion board of a computer or a function expansion unit connected to the computer, the function expansion board or the function is stored based on the instruction of the program code. The case where the CPU or the like provided in the storage unit performs part or all of the actual processing, and the above-described embodiment is realized by the processing is also included in the scope of the present invention.

As described above, according to the present embodiment, by embedding the watermark information weakly in the entire image, a method of embedding a digital watermark characterized by having a falsification position detection function becomes possible.

[0110]

As described above, according to the present invention, when data is tampered with artificially, the tampered position can be reliably detected.

Further, according to the present invention, the pixel position of the input original image data is designated, and a specific portion of the designated pixel position is rewritten based on the primary key information, so that the input image data has a predetermined value. Data can be embedded.

Further, according to the present invention, there is an effect that it is possible to detect a falsification process performed on data in which such predetermined data is embedded, and to detect the falsification position.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an outline of a digital watermark embedding device according to a first embodiment of the present invention.

FIG. 2 is a block diagram illustrating an outline of a digital watermark extraction device according to the first embodiment.

FIG. 3 is a diagram illustrating an outline of a JPEG system.

FIG. 4 is a diagram illustrating a format of data encoded by JPEG.

FIG. 5 is a diagram illustrating an example of a quantization table used in JPEG.

FIG. 6 is a diagram illustrating a zigzag scan used in JPEG.

FIG. 7 is a diagram illustrating an example for grouping AC components used in JPEG.

FIG. 8 is a diagram illustrating an example of a Huffman table of an AC component used in JPEG.

FIG. 9 is a block diagram illustrating an outline of a digital watermark embedding device according to a second embodiment of the present invention.

FIG. 10 is a block diagram illustrating an outline of a digital watermark extraction device according to a second embodiment.

FIG. 11 is a block diagram illustrating an outline of a digital watermark embedding device according to a third embodiment of the present invention.

FIG. 12 is a block diagram illustrating an outline of a digital watermark extraction device according to a third embodiment.

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5B017 AA06 BA05 BA07 BB03 CA08 CA09 CA16 5B057 CA02 CA08 CA12 CA16 CB02 CB06 CB08 CB12 CB16 CB19 CC03 CE08 CE09 CG07 CH08 CH20 DA07 DA17 5C076 AA14 AA40 BA06 CA10

Claims (26)

[Claims] An input unit for inputting image position information indicating an image position in which another data is to be embedded in the image data; a random number generating unit for generating sub-key information on a random basis based on the main key information; An image processing apparatus comprising: a pixel changing unit that changes pixel data at an image position of image data determined based on information based on the sub-key information. 2. A method according to claim 1, wherein pixel data at an image position of the image data determined based on image position information indicating an image position at which another data is embedded in the image data is randomly generated based on the main key information. Input means for inputting embedded image data obtained by changing based on information, key information generating means for inputting the main key information and generating the sub-key information, and Extracting means for extracting pixel data of the embedded image data; and comparing the pixel data extracted by the extracting means with pixel data at a pixel position based on the sub-key information generated by the key information generating means. Means for detecting a tampering position of the embedded image data according to a comparison result by the comparing means. 3. The method according to claim 1, wherein the image data is multi-valued image data, and the sub-key information is information indicating a bit position of pixel data of the multi-valued image data. The image processing apparatus according to any one of the preceding claims. 4. The image data is coded data,
The image processing apparatus according to claim 1, further comprising: an analyzing unit configured to analyze the encoded data; and a replacing unit configured to replace a code of the encoded data based on the analysis performed by the analyzing unit. 5. An encoding method wherein the encoded data is a Huffman-encoded combination of the number of coefficients from a non-zero coefficient to the next non-zero coefficient in the input sequence and the value of the next non-zero coefficient. If the data is data, the replacement means performs the Huffman code corresponding to the number of coefficients from the non-zero coefficient to the next coefficient so that the number of the Huffman code from the non-zero coefficient to the next non-zero coefficient does not change. The image processing apparatus according to claim 4, wherein the image processing apparatus is replaced with: 6. An encoding wherein the encoded data is a Huffman-encoded combination of the number of coefficients from a non-zero coefficient to the next non-zero coefficient in the input sequence and the value of the next non-zero coefficient. The image processing apparatus according to claim 4, wherein when the data is data, the replacement unit replaces a Huffman code corresponding to the value of the next coefficient. 7. The image processing apparatus according to claim 1, wherein the another data includes a digital watermark. 8. A position specifying unit for specifying a position of input image data according to embedded data, a key information generating unit for generating sub-key information from the main key information in a random manner, and a pixel at a position specified by the position specifying unit. Replacing means for replacing information with another data based on the sub-key information; calculating means for performing an operation on the data replaced with the different data by the replacing means; data replaced with the different data An image processing apparatus comprising: a synthesizing unit configured to synthesize and output a calculation result obtained by the calculation unit. 9. The pixel information at the position where the position of the input image data is specified according to the embedded data is replaced with another data based on sub-key information generated at random from the main key information and replaced with the another data. Input means for inputting data obtained by synthesizing the obtained data with the result of the operation executed, and analyzing the data input by the input means to extract the data and the result of the operation Analyzing means, calculating means for calculating the data extracted by the analyzing means, and comparing means for comparing a result of the calculation extracted by the analyzing means with a calculation result by the calculating means. Characteristic image processing device. 10. The input image data is coded data, an analyzing means for analyzing the coded data, a code position of the coded data is selected on a random basis, and a position of the selected code position is selected. 9. The image processing apparatus according to claim 8, further comprising: means for extracting bits. 11. The operation according to claim 8, wherein the operation by said operation means is an operation for calculating a hash value.
An image processing apparatus according to claim 1. 12. An input step of inputting image position information indicating an image position at which another data is embedded in image data; a random number generating step of generating sub-key information on a random basis based on main key information; A pixel changing step of changing pixel data at an image position of image data determined based on information based on the sub-key information. 13. A sub-key randomly generated based on primary key information, based on pixel data of an image position of image data determined based on image position information indicating an image position in which another data is to be embedded in the image data. An input step of inputting embedded image data obtained by making a change based on information, a key information generating step of inputting the main key information and generating the sub key information, and An extracting step of extracting pixel data of the embedded image data; and comparing the pixel data extracted in the extracting step with pixel data of a pixel position based on the sub-key information generated in the key information generating step. And a detecting step of detecting a tampering position of the embedded image data according to a result of the comparison in the comparing step. 14. The apparatus according to claim 12, wherein the image data is multi-valued image data, and the sub-key information is information indicating a bit position of pixel data of the multi-valued image data. The image processing method described in the above. 15. The image data is encoded data, further comprising: an analyzing step of analyzing the encoded data; and a replacing step of replacing a code of the encoded data based on the analysis in the analyzing step. The image processing method according to claim 12, wherein: 16. The encoded data comprises: a number of coefficients from a non-zero coefficient to a next non-zero coefficient in an input sequence;
If the next non-zero coefficient value combination is Huffman-encoded data, in the replacing step, the Huffman code corresponding to the number of coefficients from the non-zero coefficient to the next coefficient is converted to the non-Huffman code. 16. The image processing method according to claim 15, wherein replacement is performed such that the number from a zero coefficient to the next non-zero coefficient does not change. 17. The method according to claim 17, wherein the encoded data includes: a number of coefficients from a non-zero coefficient to a next non-zero coefficient in an input sequence;
In the case where the next non-zero coefficient value combination is Huffman-encoded data, in the replacement step, the Huffman code corresponding to the next coefficient value is replaced. The image processing method described in the above. 18. The image processing method according to claim 12, wherein the another data includes a digital watermark. 19. A position specifying step of specifying a position of input image data according to embedded data; a key information generating step of generating sub-key information from main key information in a random manner; and a pixel at a position specified in the position specifying step. A replacement step of replacing information with another data based on the sub-key information; an operation step of performing an operation on the data replaced by the different data in the replacement step; and the data replaced by the different data. And a combining step of combining and outputting the result of the calculation by the calculating means. 20. The pixel information at the position where the position of the input image data is specified according to the embedded data is replaced with another data based on the sub-key information generated randomly from the main key information, and replaced with the another data. An input step of inputting data obtained by synthesizing a result of the operation performed on the obtained data, and extracting the data and the result of the operation by analyzing the data input in the input step An analysis step, an operation step of operating the data extracted in the analysis step, and a comparison step of comparing a result of the operation extracted in the analysis step with an operation result of the operation step. Characteristic image processing method. 21. The input image data is coded data, wherein: an analysis step of analyzing the coded data; and randomly selecting a code position of the coded data; The image processing method according to claim 19, further comprising a step of extracting bits. 22. The image processing method according to claim 19, wherein the operation in the operation step is an operation for calculating a hash value. 23. A storage medium readable by a computer, comprising: an input step module for inputting image position information indicating an image position in which another data is to be embedded in image data; A random number generation step module that generates information, and a pixel change step module that changes pixel data of an image position of image data determined based on the image position information based on the sub-key information. Storage media. 24. A sub-key randomly generated based on main key information, based on pixel data of an image position of image data determined based on image position information indicating an image position at which another data is embedded in the image data. An input step module for inputting embedded image data obtained by changing based on information, a key information generation step module for inputting the main key information and generating the sub key information, and based on the sub key information. An extraction step module for extracting the pixel data of the embedded image data, and the pixel data extracted in the extraction step and the pixel data of the pixel position based on the sub key information generated in the key information generation step. A storage medium comprising: a comparison step module for comparing. 25. A position specifying step module for specifying a position of input image data according to embedded data, a key information generating step module for randomly generating sub-key information from main key information, and a position specified in the position specifying step. A replacement step module that replaces the pixel information with another data based on the sub-key information; an operation step module that performs an operation on the data replaced with the different data in the replacement step; A storage medium, comprising: a synthesis step module that synthesizes and outputs the replaced data and an operation result obtained by the operation means. 26. The pixel information at the position where the position of the input image data is specified according to the embedded data is replaced with another data based on the sub-key information randomly generated from the main key information, and replaced with the another data. An input step module for inputting data obtained by combining the result of the operation performed on the input data, and extracting the data and the result of the operation by analyzing the data input in the input step An analysis step module that calculates the data extracted in the analysis step; a comparison step module that compares the result of the calculation extracted in the analysis step with the calculation result in the calculation step;
A storage medium characterized by having: