JP2006270434A - Method and device for embedding information, and method and device for restoring information - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To obtain a document which is highly invisible and inseparable and is adequately restored even with respect to repeated copyings. <P>SOLUTION: When embedment information is inputted, the encode sequence group of prescribed encode sequences is generated, so as to generate encoding information, which is obtained by encoding the symbol of the embedment information through the use of the encode sequence of the encode sequence group (steps 102, 104). Then, the document data of the document, which are obtained by embedding the embedment information in the image, is generated by acquiring two-dimensional information to be an even function, when the encoding information is arranged on a two-dimensional plane, generating an embedded information image by performing inverse-Fourier transformation with respect to the two-dimensional information, and performing compositing with the image of an object document (steps 106-110). <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an information embedding method, an information embedding device, an information restoration method for restoring embedded information, and an information restoration device for embedding predetermined information in a document on which an image is formed.

  With the spread of computerization of various information, leakage of confidential information such as customer lists and product information under development has become a problem. In particular, digitized documents (document data) can be easily duplicated exactly as the original, and taking out duplicated document data is extremely easy by using a network, etc. The prevention of document data leakage is desired.

  In recent years, encryption technology, authentication technology, access restriction technology, etc. have made rapid progress, and taking appropriate measures using these technologies has made it difficult to take out document data.

  On the other hand, with the widespread use of printers and copiers and advanced functions, not only print output of document data but also elaborate document copy is possible, and paper documents created by print output or copy are It is extremely easy to take out.

  From this, the prevention of information leakage due to the electronic document data being taken out as a paper document is delayed.

  By the way, there is a technique for adding tracking information as a security measure for confidential information on paper documents. This tracking information includes information that can specify the output person when print output is performed, information that specifies the output date and time, and the like. By recording this tracking information on the paper document, the paper document When the information leaks to the outside, it is possible to specify the leak source from the tracking information, so that leakage of confidential information can be suppressed.

  Such tracking information requires invisibility when formed on a paper document and inseparability from content such as images recorded on the paper document.

  As a method of recording information on a paper document, one-dimensional or two-dimensional barcodes are generally used, but tracking information is visible or recorded by symbols unrelated to the content on the paper document. If this is done, it will become clear that the tracking information has been recorded. For this reason, it is possible to delete the information by cutting or painting it, and to decode or alter the information. Since the original identification is disturbed, tracking information is required to be invisible or inseparable from the content.

  From this, especially in paper area documents that include images, in the area gradation processing that is performed when printing a multi-tone image, the screen angle is switched according to the information, so that information can be added to the image formed on the document. A method of embedding has been proposed (see, for example, Patent Document 1).

  In this proposal, when dividing an image into a plurality of blocks and embedding 1-bit information for each block, by representing bit 0 as a screen angle of 0 degrees and expressing bit 1 as a screen angle of 45 degrees, Information is embedded in the print image.

  In addition, a proposal has been made to embed corresponding information by arranging information that identifies the device that performed the image output and the operator who performed the operation as a yellow dot pattern on the paper at regular intervals. (For example, refer to Patent Document 2).

  In this proposal, the invisibility of embedded information is enhanced by forming a pattern indicating information with fine dots and using yellow, which is a color that is difficult to visually recognize.

  Furthermore, a proposal has been made to embed information by multiplying the information to be added by a pseudo-noise sequence to disperse and transform the spatial spectrum of the information information on the original image, and adding this to the original image (for example, patents). Reference 3).

  In this proposal, not only is it possible to embed information without being limited to multicolor images and single color images, but also the power spectral density of the information to be added can be reduced. It can be kept small.

  On the other hand, as the performance of copying machines improves, highly reproducible image copying is possible, and paper documents are easily copied by copying machines, so they are taken out as secondary and tertiary copies. Sometimes. For this reason, when the tracking information is embedded in a paper document, resistance to repeated copying (copy resistance) is required.

  In addition, when copying is repeated, image quality deterioration occurs due to a significant change in density, attenuation of high-frequency components of the image, noise or roughness on the image, etc. Tolerance to is also required.

  However, in the proposal of Patent Document 1, if the high frequency component is lost due to repeated copying, the fine structure of the screen is lost. As a result, it is difficult to detect the screen angle from a paper document generated by repeated copying, and it is difficult to restore the embedded information.

  According to the proposal of Patent Document 2, the yellow dot pattern disappears when copied by a monochrome copying machine. Even if the dot pattern is printed in black, if a minute dot is used, the high frequency component of the image is lost due to repeated copying, which makes it difficult to restore the embedded information. In addition, if a large dot pattern is used to prevent this, the invisibility is lowered.

Furthermore, in the proposal of Patent Document 3, since spatial spectrum dispersion processing is performed on the embedded information, the converted information has a frequency component much higher than that of the original embedded information. For this reason, if the high frequency component of the image is lost due to repeated copying, the high frequency component of the converted information is greatly lost, and it becomes extremely difficult to restore the embedded information.
JP 2002-223346 A Japanese Patent No. 3367959 Japanese Patent No. 3251385

  The present invention has been made in view of the above facts, and can obtain high inseparability with respect to contents such as images and can maintain high invisibility on a paper document, so that accurate information can be obtained even when copying is repeated. It is an object to provide an information embedding method, an information embedding device, an information restoring method for restoring embedded information, and an information restoring device.

  In order to achieve the above object, an information embedding method of the present invention is an information embedding method for embedding predetermined information for a document on which an image is formed as an information embedding target, and each symbol based on the embedding information is encoded with a predetermined code The encoded information encoded by the sequence group generates two-dimensional information arranged so as to be an even function on the two-dimensional plane, and the two-dimensional information is subjected to inverse Fourier transform to thereby generate an image of the document to be embedded. An information image to be embedded is formed.

  According to the present invention, a symbol (symbol value) corresponding to each data value of the embedded information is encoded by the encoded sequence group formed by a predetermined encoded sequence to generate encoded information, and this encoded information 2D information that is an even function on the 2D plane is generated. An embedded information image is formed by performing inverse Fourier transform on the two-dimensional information.

  By combining the embedded information image formed in this way with the image of the target document, it is possible to embed information with high invisibility and inseparability.

  The information embedding method of the present invention is characterized in that the encoded information is arranged in an intermediate frequency region on the two-dimensional plane that is a frequency plane.

  According to the present invention, an embedded information image is formed except for a low frequency region and a high frequency region that are easily affected by image quality degradation.

  Thereby, it is possible to improve the reconstructability of embedded information even for a document that has been repeatedly copied.

  In the present invention, it is preferable that the two-dimensional information is generated by multiplying a coefficient corresponding to the arrangement position for each of the encoded information on the two-dimensional plane.

  In the present invention, the code sequence groups are preferably orthogonal or substantially orthogonal to each other and are random code sequences. Further, for each code sequence in the code sequence group, It is preferable that one symbol of the embedded information is allocated.

  On the other hand, in the information restoration method for the information embedding method of the present invention, the encoded information obtained by encoding each symbol based on the embedded information with a predetermined code sequence group is arranged so as to be an even function on a two-dimensional plane. An information restoration method for restoring the embedding information from a document or document data obtained by synthesizing an embedding information image generated by inverse Fourier transform of dimensional information, the image data of the document or document data being Fourier transformed Obtaining a frequency space representation, extracting a real number component serially from the frequency space representation to generate an extracted information sequence, and obtaining a correlation value between the extracted information sequence and a code sequence in the code sequence group, The embedded information is restored.

  According to the present invention, the frequency space representation is obtained by Fourier transforming the image data of the document whose information is to be restored, and the real number components are extracted in series by this frequency space representation, so that the extraction information corresponding to the embedded information is obtained. Generate a series.

  By calculating a correlation value between the extracted information sequence and the code sequence of the code sequence group used for information embedding, the embedded symbol is determined, and the embedded information is restored.

  In such an information restoration method of the present invention, the extracted information series is generated by using the real number components extracted in series with a coefficient corresponding to the arrangement position of the image data on the frequency space representation. be able to.

  Such an information embedding device applicable to the present invention is an information embedding device for embedding predetermined information for a document on which an image is formed as an information embedding target, and for inputting document data of the document to be embedded Means for generating a code sequence group based on a predetermined code sequence, encoding each symbol forming the embedded information by the generated code sequence to generate an encoded information sequence, and the encoding Two-dimensional information generating means for generating two-dimensional information arranged so that series information is an even function on a two-dimensional plane, and an embedded information image corresponding to the embedded information by performing inverse Fourier transform on the two-dimensional information First conversion means for generating the information, and information synthesis means for synthesizing the embedded information image with the image of the document to be embedded. If good, the with embedded information is input, as long as it contains an embedded information input means to replace the symbol which is previously set embedded information entered.

  The information restoration apparatus applied to the present invention is a two-dimensional arrangement in which encoded information obtained by encoding each symbol based on embedded information with a predetermined code sequence group is an even function on a two-dimensional plane. An information restoration apparatus for restoring the embedded information from a document or document data obtained by synthesizing an embedded information image generated by performing inverse Fourier transform on information, and for reading image data from the document or document data And a second transforming means for performing a Fourier transform on the image data to generate a frequency space representation of the image data, and an extraction for generating an extracted information sequence by serially extracting real components from the frequency space representation An information sequence generating means calculates a correlation value of each code sequence of the extracted information sequence and the code sequence group, and determines each symbol of the embedded information from the calculation result. And embedding information restoring means for restoring the embedded information by, as long as it contains a.

  Furthermore, the present invention may use a program that can achieve the above functions.

  As described above, according to the present invention, embedded information symbols can be encoded with a predetermined code sequence group and embedded with extremely low embedding strength, so that high invisibility can be obtained and an image in a document can be obtained. Therefore, it is possible to embed the embedded information on the image data on the document data or on the image of the document printed based on the document data so that it is difficult to alter or remove the information.

  At the same time, since embedded information can be embedded in a frequency region where image degradation does not occur, even if image quality degradation due to repeated copying occurs in an image on a document, the embedded information is not affected by the image quality degradation. . Furthermore, as a result of the information restoration by the code sequence group, components other than the embedded information subjected to the encoding process are greatly reduced, so that an excellent effect that an accurate restoration property can be maintained is obtained.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 shows a schematic configuration of a document processing system 10 applied to the present embodiment. The document processing system 10 includes a document processing device 12 that can add predetermined information when printing a document based on document data including an image (image data), and a document processing device to which the predetermined information is added. 12 includes a document processing apparatus 16 that extracts predetermined information from a document 14 printed out from the document 12 or a document 14A obtained by copying the document 14.

  The document processing devices 12 and 16 include information processing devices 18 and 20 such as personal computers and workstations. The document processing apparatus 12 includes a printer 22 that performs a printing process based on image data (document data) output from the information processing apparatus 18, and the document processing apparatus 16 converts the documents 14, 14A and the like into originals. The scanner 24 is configured to read an image recorded on a document as an image and output image data (document data) corresponding to the image to the information processing apparatus 20.

  The document processing system 10 may be capable of exchanging data by connecting the document processing apparatus 12 and the document processing apparatus 16 to a network. Although the description will be divided into the document processing devices 12 and 16 here, it may be configured using a single document processing device having the functions of the document processing devices 12 and 16. That is, the printer 22 and the scanner 24 may be connected to one information processing apparatus. At this time, any configuration can be applied, such as using a copier or a multifunction machine having both a printer function and a scanner function.

  FIG. 2 shows a functional block diagram of the document processing apparatus 12. The document processing apparatus 12 includes an image input unit 26 and an embedded information input unit 28.

  In the document processing system 10, when printing the document 14 from the document processing device 12, it is possible to embed arbitrary information such as information such as the print output date and time, the print output person, and the printer that has performed the print (hereinafter referred to as “print output”) The image input unit 26 receives image data of a document to be embedded, and the embedded information input unit 28 prints image data or a document that is printed out according to the image data. 14 embedded information to be embedded.

  As image data, image data inserted into electronic document data in a page description language such as PostScript or word processor output format, or a raster of an image read or photographed by an image reading means such as a scanner or a digital camera or an imaging means. It may be data.

  The image input unit 26 can include a scanner, a digital camera, and the like, and can store and store image data such as a hard disk drive (HDD), a DVD (Digital Versatile Disc, a DVD-RAM, a DVD ± RW). , DVD ± R etc.) drive, CD (Compact Disc, CD-ROM, CD-R, CD-RW etc.) drive, MO (Magneto-Optical disk) drive etc. Any configuration capable of receiving image data can be included, such as a data transfer device that transmits and receives image data via a network.

  The embedded information input from the embedded information input unit 28 may be any data as long as it is digital data or data that can be converted into digital data, such as numbers, characters, URL (Uniform Resource Locator), multimedia data such as sound and images. Can be applied.

  As the embedded information input unit 28 for inputting such embedded information, characters and the like can be input using an input device such as a keyboard. When inputting multimedia data, the same as the image input unit 26 described above. In addition, a configuration in which data is exchanged through various storage media such as HD, DVD, CD, and MO and a network can be applied.

  Further, the document processing apparatus 12 is provided with an information composition unit 30, and image data such as electronic image data and raster data is input from the image input unit 26 to the information composition unit 30.

  On the other hand, the document processing apparatus 12 includes a code sequence group generation unit 32, an information encoding unit 34, a two-dimensional information generation unit 36, and an inverse Fourier transform unit 38.

  The code sequence group generation unit 32 generates one or more code sequences to be applied to generation of encoded information obtained by encoding the embedded information input from the embedded information input unit 28. At this time, the code sequence group generation unit 32 generates a number of code sequences corresponding to the number of symbols of the embedded information as a code sequence group.

  The information encoding unit 34 encodes the embedded information input from the embedded information input unit 28 with the code sequence group generated by the code sequence group generation unit 32, and generates the encoded information sequence corresponding to the embedded information and the code sequence group. Generate. At this time, the information encoding unit 34 encodes each symbol of the embedded information using a code sequence in the code sequence group corresponding to the corresponding symbol, and synthesizes the encoding result of each symbol to encode the encoded information sequence. Is generated.

  The two-dimensional information generation unit 36 generates two-dimensional information for arranging the encoded information sequence generated by the information encoding unit 34 at a predetermined position on the two-dimensional plane. At this time, the two-dimensional information generation unit 36 arranges the encoded information sequence so that the generated two-dimensional information is an even function. At this time, the strength of each symbol of the encoded information sequence may be changed by multiplying by a coefficient corresponding to the position on the two-dimensional plane.

  The inverse Fourier transform unit 38 obtains a real space representation of the embedded information image by performing inverse Fourier transform on the two-dimensional information generated by the two-dimensional information generation unit 36.

  The real time representation of the embedded information image generated by the inverse Fourier transform unit 38 is input to the information synthesis unit 30 together with the image data of the target document. At this time, the information synthesizing unit 30 generates image data in which additional information is embedded in the image data of the target document by synthesizing the image data of the target document and the real space representation of the embedded information image.

  The information synthesizing unit 30, the code sequence group generating unit 32, the information encoding unit 34, the two-dimensional information generating unit 36, and the inverse Fourier transform unit 38 are a CPU that performs various processes, and a processing program that is executed by the CPU. It can be configured using hardware of a general configuration in a personal computer, a workstation, or the like, such as a stored ROM, RAM or HDD used as a temporary storage means, or the like.

  Further, the document processing apparatus 12 is formed with an image output unit 40 that executes a printing process or the like using the printer 22. The image output unit 40 converts the image data combined with the embedded information by the information combining unit 30 into a predetermined format and outputs it. At this time, the image output unit 40 converts the image data into a data format that can be processed by the printer 22, and outputs the data to the printer 22, thereby printing the image data in which predetermined embedding information is embedded on paper. Document 14 can be obtained.

  The image output unit 40 is not limited to print output on paper, and is output as, for example, image data inserted into electronic document data in a page description language such as PostScript or word processor output format, or a data file in raster data format. It may be what you do.

  Such an image output unit 40 is not only a printer 22 or a control device that controls the printer 22, but also a data writing device that enables writing of data to various recording media, a control device for the data writing device, and a network. A data transfer device that transmits data can be connected.

  On the other hand, FIG. 3 shows a functional block diagram of the document processing apparatus 16. The document processing device 16 is provided with an image reading unit 42. The image reading unit 42 reads image data such as raster data of a document in which embedded information is embedded. For example, when the target document is a document 14 printed out on paper or a document 14A created by copying the document 14, the documents 14 and 14A are loaded into the scanner 24 and the image is read. Then, raster data corresponding to the documents 14 and 14A is generated, and the raster data is output as image data.

  As such an image reading unit 42, not only an image reading unit such as the scanner 24 but also an imaging unit such as a digital camera can be used. Further, the image reading unit 42 can include a data reading device corresponding to the storage medium when the target document is recorded on the storage medium as electronic image data or raster data, or via a network. It may include a configuration for reading data. Note that the image reading unit 42 can also have a function of converting image data input in various data formats into raster data.

  Further, the document processing device 16 includes a Fourier transform unit 44, an information sequence extraction unit 46, a code sequence group generation unit 48, and an embedded information restoration unit 50.

  The Fourier transform unit 44 performs a Fourier transform on the image data input from the image reading unit 42. That is, the Fourier transform unit 44 obtains a frequency space representation of the image in which the additional information is embedded by performing Fourier transform on the image data in which the additional information is embedded.

  The information sequence extraction unit 46 extracts an encoded information sequence from the frequency space representation of the image obtained by the Fourier transform unit 44. That is, the information sequence extraction unit 46 extracts an embedded encoded information sequence from the frequency space representation of an image in which additional information is embedded. At this time, the encoded information sequence extracted by the information sequence extraction unit 46 includes not only embedded information but also an image component, and when the document 14A has been subjected to repeated copying, a deteriorated component due to repeated copying. Is also included.

  On the other hand, in the document processing device 16, the embedded information embedded in the document by the document processing device 12 is restored. From here, the code sequence group generation unit 48 includes the code sequence group generation unit of the document processing device 12. The same code sequence group as the code sequence group generated in 32 is generated.

  As such a code sequence group generation unit 48, as shown in FIG. 2, a code sequence group storage unit 52 is provided in the document processing apparatus 12 to store a code sequence group used when embedding additional information in a document. The code sequence group stored in the code sequence group storage unit 52 may be read. At this time, as shown in FIG. 1, the document processing apparatus 12 and the document processing apparatus 16 are connected via a network. Also good.

  In addition, the code sequence group may be set in advance and stored in each of the document processing device 12 and the document processing device 16, and the code sequence group may be stored in a predetermined storage medium. The code sequence group generation unit 48 may read from the corresponding storage medium, but it is preferable to apply a method that ensures security.

  The embedded information restoration unit 50 (see FIG. 3) restores the embedded information from the information encoded sequence extracted by the information sequence extraction unit 46 using the code sequence group generated by the code sequence group generation unit 48.

  The Fourier transform unit 44, the information sequence extraction unit 46, the code sequence group generation unit 48, and the embedded information restoration unit 50 include a CPU that performs various processes, a ROM that stores a processing program executed by the CPU, and a temporary storage unit. It can be configured using hardware of a general configuration in personal computers, workstations, and the like, such as RAM and HDD used.

  In addition, the document processing device 16 is provided with an information output unit 54 that outputs the embedded information restored by the embedded information restoring unit 50. The information output unit 54 may display embedded information using various display devices such as a CRT and an LCD, and may print out using a print output device such as a printer. It is possible to adopt any configuration such as outputting the embedded information by writing it in a storage medium, as long as it includes control means or an output device corresponding to the output destination and output format.

  Here, a description will be given of a process for creating the document 14 in which the predetermined embedding information is embedded, and restoring the embedding information from the document 14 or the document 14A created by copying the document 14 (hereinafter referred to as the document 14A). To do.

  FIG. 4 shows an outline of a document 14 creation process in the document processing apparatus 12 as an example of embedding additional information. In this flowchart, in the first step 100, image data in which additional information is embedded (image data of the target document) is input.

  In step 102, embedded information is input. When the embedded information is input, in step 104, the embedded information is encoded to generate a code sequence group.

  Here, for example, if the information to be embedded is binary data of n bits (the number of bits n), in step 102, the embedded information to be input is converted into binary data of n bits, and the embedded information b (i) is converted into the embedded information b (i). Generate. The image data input in step 100 is f (x, y) and the size is X × Y.

At this time, it is assumed that “−1” is assigned as a data value to a bit whose data value is “0”. That is, the symbol value of information takes any of ± 1. This
b (i) = {± 1} (where i = 0, 1, 2,..., n−1)
Is obtained.

  In addition, as the code sequence, a sequence of values having a small average value of symbol values and having randomness is used. The generated code sequence group is such that the absolute value of the cross-correlation value is small between any two code sequences.

  Examples of such a code sequence include an m-sequence (maximum-length sequence) and a Gold code (Gold sequence), and these can be used. Any code sequence group can be used as long as it satisfies the requirements.

Here, as an example of the code sequence group, when an m sequence having a sequence length L of one period and a shift version thereof are used, the code sequence group pn (i, k) generated by the code sequence generation unit 32 is
pn (i, k) = m (k + i)
(Where k = 0, 1, 2,..., L-1, i = 0, 1, 2,..., N-1)
It becomes. At this time, the code sequence m (x) represents an m-sequence having a sequence length L of one cycle, and the symbol value thereof takes either “+1” or “−1”.

  In step 104 of FIG. 4, a coding process is performed on each bit of the embedded information b (i) using the code sequence group pn (i, k).

  FIG. 5 shows an outline of the encoding process. Here, it is assumed that the encoded information sequence c (k) shown in Equation 1 (1) is generated by the encoding process.

  From the flowchart shown in FIG. 5, the initial setting (i = 0) is performed in the first step 120, and the embedded information b (i) is encoded by the code sequence group pn (i, k) in the next step 122. To do. Thereafter, in step 124, the encoded result is added to the code sequence c (k).

  In step 126, it is confirmed whether i has not reached n of the number of bits. If it has not reached, an affirmative determination is made in step 126, the process proceeds to step 128, i is incremented, and the process returns to step 122. By repeating this, a coded information sequence c (k) based on n pieces of coded information is generated.

  Here, since both the embedded information b (i) and the code sequence pn (i, k) have symbol values “+1” or “−1”, their products are also “+1” or “−1”. . Accordingly, the encoded information sequence c (k) is a sequence having symbol values in the range [−n, n].

  In the flowchart of FIG. 4, when the embedded information is encoded and the encoded information sequence c (k) is generated, the process proceeds to step 106, and the encoded information sequence c (k) is arranged on the two-dimensional plane. Dimension information C (u, v) is generated. The two-dimensional information C (u, v) is a frequency space representation of an embedded information image to be synthesized with the image data f (x, y) of the target document. The size of the two-dimensional information C (u, v) is N × M.

  In the two-dimensional information C (u, v), the encoded information sequence c (k) is arranged on a two-dimensional plane. FIG. 6 shows an outline of a two-dimensional plane having the point O as the coordinate origin. In the two-dimensional plane of FIG. 6, the frequency is lowest at the coordinate origin O, and the frequency is high in the horizontal direction (N) direction and the vertical direction (M direction).

  Here, if the region where the encoded sequence information c (k) is arranged is the region 60A, and the region for converting the two-dimensional information C (u, v) into an even function is the region 60B, the regions 60A and 60B are excluded. The regions 60C and 60D are zero.

  At this time, the region 60A is preferably set only in the intermediate frequency region.

  In other words, remarkable image quality degradation due to repeated copying is a significant change in copy density, attenuation of high-frequency components of the image, or noise and roughness on the image. Appears as fluctuations in the low frequency component of the image, attenuation, noise, and roughness of the image appear mainly in the high frequency component of the image, and the intermediate frequency range is relatively stable (compared to the low frequency range and high frequency range). ing. From this point, by placing the encoded information sequence c (k) in the region 60A, which is the intermediate frequency region, it is possible to embed information that is robust to repeated copying.

  Here, the encoded sequence information c (k) is arranged at a predetermined position on the two-dimensional plane. It should be noted that the position in the area 60A where the encoded sequence information c (k) is arranged does not require any particular rule so that the encoded sequence information c (k) can be extracted in the same order later when the encoded sequence information c (k) is restored. What is necessary is just to determine the embedding place for each k.

  The two-dimensional information C (u, v) for the encoded sequence information c (k) at this time is as follows.

C (u (k), v (k)) = c (k)
At this time, the encoded sequence information c (k) is simply arranged on the two-dimensional plane, but may be multiplied by a coefficient W (u, v) corresponding to the position on the two-dimensional plane. .

C (u (k), v (k)) = W (u (k), v (k)) · c (k)
Thereby, the strength of the symbol of the encoded information sequence c (k) can be changed.

  In general, in a natural image, power is concentrated in a low-frequency region, and the image power rapidly decreases as the spatial frequency of the image increases. In the present invention, the encoded information sequence c (k) is embedded in such an image.

  However, when embedding information is restored, there is no problem because the information sequence arranged in the high frequency region has less image power, but the information sequence arranged in the low frequency region adds a strong image power. It may be a noise component and hinder the restoration of embedded information.

  Therefore, if the strength of the information series arranged in the low frequency region is increased in advance, the influence of the image can be reduced when embedding information is restored.

On the other hand, a region 60B shown in FIG. 6 is a region where data for converting the two-dimensional information C (u, v) into an even function is arranged, and the two-dimensional information C (u, v) in the region 60A. In between,
C (N−u, M−v) = C (u, v)
Have a relationship.

  Since the two-dimensional information C (u, v) is a frequency space representation of the embedded information image, an image obtained by performing inverse Fourier transform on the embedded information image is an image synthesized with the image of the target document. However, since the inverse Fourier transform is a transform for a complex signal, even if the two-dimensional information C (u, v) is a real function, the inverse Fourier transform result is generally a complex function. Further, the pixel value of the image must be a real number, and the inverse Fourier transform of the two-dimensional information C (u, v) requires a process that becomes a real function, and the process at this time is an even function.

  That is, since the inverse Fourier transform of the real even function becomes a real even function, it can be directly synthesized with the image of the target document.

  In step 108 of FIG. 4, the inverse Fourier transform is executed to generate an embedded information image d (x, y) obtained by performing inverse Fourier transform on the two-dimensional information C (u, v). At this time, since the two-dimensional information C (u, v) is a real even function, the embedded information image d (x, y) obtained by performing inverse Fourier transform on this has a real value. Equation 2 shows the inverse Fourier transform at this time.

In the next step 110, an embedded information image d (x, y) is synthesized with the input image f (x, y), and an image (image data) g (x) in which the embedded information is embedded in the image data of the target image. , Y). That is,
g (x, y) = f (x, y) + δ · d (x, y)
Execute.

  Here, the constant δ is an information embedding strength that takes a positive value, and in order to increase invisibility and robustness, the constant δ is preferably set to the largest value in a range where embedding information is not noticeable. .

  Here, it is assumed that the size of the input image X × Y is equal to the size of the embedded information image N × M. However, when the embedded information image is smaller, the embedded information image is It may be arranged at a specific position on the input image, or may be arranged repeatedly in a tile shape.

  When the information images are combined in this way, in the next step 112, a printing process based on the image data of the combined image is performed, and the document 14 in which predetermined additional information is embedded is printed out.

  In addition, when outputting as document data, what is necessary is just to output to the designated output destination by converting into a predetermined data format (format).

  In this way, the document 14 in which the additional information is embedded is inseparable in the image that is the content on the document 14, unlike the document 14 that is written together with the content such as text in the document. Embedded in format.

  As a result, if the embedded information is to be removed from the document 14, the content itself in the document 14 is lost, and only the embedded information cannot be removed without damaging the content in the document 14.

  Therefore, high security can be ensured by applying tracking information for a document that is not desired to be leaked to the outside as embedded information. That is, since it is difficult to remove the embedded information, the origin of the document 14 can be clarified from the information embedded in the document 14.

  Next, restoration of embedded information from document 14 or document 14A created by copying document 14 (hereinafter referred to as document 14A) will be described.

  The document processing device 16 is used to restore the embedded information from the document 14A. FIG. 7 shows the flow of processing in the document processing device 16 at this time. In this flowchart, in the first step 130, the image data of the document 14A is read. This image data can be read by applying a method such as loading the document 14A into the scanner 24 and reading it.

  The reading of the image data of the document 14A is not limited to the image reading unit such as the scanner 24, and an imaging unit such as a digital camera can be used. The target for restoring the embedded information is not only the document 14A but also the image data in which the information is embedded (image data that can output the document 14). At this time, the target image data is stored. Read via media or network.

Here, there is a possibility that the read image data includes image deterioration due to repetitive copying. From here, the read image (image data) g (x, y) is compared with the read image (image data). If the data) is an image g ′ (x, y), the image g ′ (x, y) is
g ′ (x, y) = f (x, y) + δ · d (x, y) + e (x, y)
It can be expressed as. At this time, the image g ′ (x, y) includes an error component when the document 14 is printed out, an image quality deterioration component due to repeated copying, and an error component generated when the image is read. From here, the sum of the error component and the degradation component is defined as an error component e (x, y).

  In the next step 132, the image g ′ (x, y) is subjected to Fourier transform to obtain a frequency space representation G ′ (u, v) of the image g ′ (x, y). At this time, since the Fourier transform is a linear transform, G (u, v) can be expressed by Equation 3.

  In Equation 3, F (u, v) is the Fourier transform of the original image f (x, y) to be embedded, and E (u, v) is the Fourier transform of the error component e (x, y). Represents. Here, it is assumed that the size X × Y of the image g ′ (x, y) is equal to the size N × M of the embedded information image, but when the embedded information image is smaller, the image G ′ (u, v) may be obtained by cutting out one section of the embedded information image from g ′ (x, y) and performing Fourier transform.

  In the next step 134, the real number component at the arrangement position of the encoded information sequence is extracted from the frequency space representation G ′ (u, v) obtained by Fourier transform of the image g ′ (x, y).

The encoded information sequence c (k) used when embedding the embedded information image is
C (u (k), v (k)) = c (k)
And the extracted information series c ′ (k),
c ′ (k) = Re (G ′ (u (k), v (k))
= F (k) + δ · c (k) + e (k)
It becomes.

Here, if Re (x) represents the real component of the complex number x, f (k) and e (k) are respectively
f (k) = Re (F (u (k), v (k)))
e (k) = Re (E (u (k), v (k)))
It becomes. When the symbol intensity of the encoded information sequence c (k) is changed by a constant W (u, v), the extracted information sequence c ′ (k) may be set as follows.

c ′ (k) = Re (G ′ (u (k), v (k))) / W (u (k), v (k))
= F (k) / W (u (k), v (k)) + δ · c (k) + e (k) / W (u (k), v (k))
Here, f (k) / W (u (k), v (k)) and e (k) / W (u (k), v (k)) are expressed as f (k) and e (k). In other words, it can be expressed in the same way as c (k) = f (k) + δ · c (k) + e (k).

  In the next step 136, a code sequence group is generated. At this time, the same code sequence group pn (i, k) as that applied to information embedding in the document 14 is generated. Note that instead of generating the code sequence group pn (i, k), the code sequence group pn (i, k) may be acquired from the document processing device 12 or the like. In addition, the code sequence group pn (i, k) applied to the creation of the document 14 can be obtained reliably.

  Thereafter, in step 138, the embedded information is restored. FIG. 8 shows an outline of restoration processing of embedded information. In this flowchart, initial setting (i = 0) is performed in the first step 150. Thereafter, in step 152, a correlation value R (i) between the extracted information sequence c ′ (k) and the code sequence group pn (i, k) is calculated.

  Here, the correlation value R (i) between the extracted information sequence c ′ (k) and the i-th code sequence pn (i, k) can be calculated by Equation 4.

  The correlation value R (i) can be rewritten as shown in Equation 1 to Equation 5. From here, the correlation value R (i) can be rewritten as shown in Equation 5.

  Here, paying attention to the second term in the last row of the above equation, the code sequence group pn (i, k) is a code sequence group having a small absolute value of the cross-correlation value regardless of any two code sequences in the group. The second term takes a large value only when i = j, and is otherwise very small compared to when i = j.

  That is, in the present embodiment, an m-sequence and its shifted version are applied as an example of the code sequence group pn (i, k), and it is well known that the cross-correlation value between them is expressed by Equation 6.

  In addition, since the sequence length L of the code sequence is several hundred to thousands, the cross-correlation value of Equation 6 takes a very large value only when i = j. From this, the equation shown in Equation 6 can be rewritten as shown in Equation 7.

  Here, as the sequence length L of the code sequence, a large value (hundreds to thousands) is usually used. Therefore, if the value of the constant δ of the information embedding strength is appropriately set, The absolute value of the two terms is a very large value.

  On the other hand, the first term and the fourth term of the same equation are respectively the cross-correlation value of the image component series f (k) and the code series pn (i, k) and the error component series e (k). And the code sequence pn (i, k) are expressed as cross-correlation values, and their absolute values are much smaller than the second term of the same equation. That is, since the code sequence pn (i, k) has a small average value of symbol values and has randomness, it has low correlation with any signal other than itself (correlation). Low value).

  The absolute value of the third term of the equation is (n−1) · δ or less, and if the number of bits n of information to be embedded is a value sufficiently smaller than the sequence length L of the code sequence, The absolute value of the third term is much smaller than the absolute value of the second term.

  From this point, if the sequence length L of the code sequence is sufficiently large and the number of bits n of information to be embedded is sufficiently small compared to the sequence length L, the absolute value of the second term is the absolute value of the other term. Therefore, the correlation value R (i) can be approximated to the second term of the last row of Equation 5 (the second term of the last row of Equation 6), as shown in Equation 8.

  In Equation 8, since the information embedding strength δ and the sequence length L of the code sequence are both positive values, the code for each bit of the embedded information can be determined from the code of the correlation value R (i).

  From this point, in the flowchart of FIG. 8, when the correlation value R (i) is calculated in step 152, the sign of the calculated correlation value R (i) is confirmed in step 154. At this time, if the correlation value R (i) is positive, an affirmative determination is made in step 154 and the process proceeds to step 156 where the corresponding embedded information b (i) is set to “1” and the correlation value R (i ) Is negative, a negative determination is made in step 154, the process proceeds to step 158, and the corresponding embedded information b (i) is set to “−1”.

  In step 160, it is confirmed whether i has not reached n. If it has not reached, affirmative determination is made in step 160, the process proceeds to step 162, i is incremented, and the process returns to step 152.

  Thereby, the embedded information b (i) is restored.

  When the embedding information is restored in this way, in the flowchart of FIG. 7, the process proceeds to step 140 and the embedding information is output. The output method at this time may be print processing, may be displayed on a display device such as a CRT or LCD, and is stored in various storage media as digital data of embedded information. In addition, an output method such as forwarding to a predetermined destination via a network or the like can be applied.

  As described above, the document processing apparatus 16 can accurately restore the embedded information from the document data of the document 14 or the document 14 created by embedding predetermined information in the image by the document processing apparatus 12.

  Further, since the document 14 is repeatedly copied and created, even if a significant change in image density, attenuation of high-frequency components of the image, or deterioration of image quality such as noise or roughness on the image occurs, it is substantially affected. An embedded information image is generated using an intermediate frequency region that is not subject to the reception. Furthermore, as a result of the information restoration process using the code sequence group, components other than the embedded information subjected to the encoding process are significantly reduced, so that the embedded information is also obtained from the document 14A in which the image quality has deteriorated. Can be accurately restored.

  The present embodiment described above shows an example of the present invention and does not limit the configuration of the present invention. For example, in the present embodiment, the document processing system 10 configured by the document processing apparatuses 12 and 16 has been described as an example. However, the present invention uses an information processing apparatus having an arbitrary configuration, It can be configured by incorporating hardware or software (program).

It is a schematic block diagram of the document processing system applied to this Embodiment. It is a block diagram which shows schematic structure of the principal part of the document processing apparatus used for the document preparation which embedded information. It is a block diagram which shows schematic structure of the principal part of the document processing apparatus used for information restoration. It is a flowchart which shows the outline of document preparation. It is a flowchart which shows the outline of the encoding process of embedded information. It is the schematic which shows the area | region which arrange | positions the code sequence group on a frequency plane. It is a flowchart which shows the outline of a process when embedding information is decompress | restored. It is a flowchart which shows the outline of a process when restoring embedding information.

Explanation of symbols

10 Document Processing System 12 Document Processing Device (Information Embedding Device)
14 document 14A document 14B document 16 document processing device (information restoration device)
18 Information processing device (information embedding device)
20 Information processing device (information restoration device)
DESCRIPTION OF SYMBOLS 22 Printer 24 Scanner 26 Image document input part 28 Embedded information input part 30 Information synthetic | combination part 32 Code sequence group production | generation part 34 Information coding part (encoded information series production | generation part)
36 Two-dimensional information generation unit 38 Inverse Fourier transform unit (first transform unit)
40 Image output unit 42 Image reading unit 44 Fourier transform unit (second transforming means)
46 Information sequence extraction unit (extraction information sequence generation means)
48 Code sequence group generation unit 50 Embedded information restoration unit (information restoration unit)
54 Information Output Unit 52 Information Output Unit

Claims (10)

An information embedding method for embedding predetermined information for a document on which an image is formed as an information embedding target,
Two-dimensional information in which the encoded information obtained by encoding each symbol based on the embedded information with a predetermined code sequence group is an even function on a two-dimensional plane is generated,
An information image to be embedded in the image of the document to be embedded is formed by performing an inverse Fourier transform on the two-dimensional information.
An information embedding method characterized by the above.   The information embedding method according to claim 1, wherein the encoded information is arranged in an intermediate frequency region on the two-dimensional plane that is a frequency plane.   3. The information embedding method according to claim 1, wherein the two-dimensional information is generated by multiplying a coefficient corresponding to an arrangement position for each of the encoded information on the two-dimensional plane.   4. The information embedding method according to claim 1, wherein the code sequence group is a code sequence that is orthogonal or substantially orthogonal to each other and has randomness. 5.   5. The information embedding method according to claim 1, wherein one symbol of the embedding information is allocated to each code sequence in the code sequence group. Embedding information generated by performing inverse Fourier transform on two-dimensional information in which each symbol based on the embedding information is encoded with a predetermined code sequence group so that the encoded information becomes an even function on the two-dimensional plane. An information restoration method for restoring the embedded information from a document or document data obtained by combining images,
The image data of the document or document data is subjected to Fourier transform to obtain a frequency space expression, and a real number component is extracted in series from the frequency space expression to generate an extracted information series,
Reconstructing the embedded information by obtaining a correlation value between the extracted information sequence and a code sequence in the code sequence group;
An information restoration method characterized by the above.   7. The information according to claim 6, wherein the extracted information series is generated in accordance with the real number component extracted in series by a coefficient corresponding to an arrangement position of the image data on the frequency space expression. Restoration method. An information embedding device that embeds predetermined information on a document on which an image is formed as an information embedding target,
Image input means for inputting document data of the document to be embedded;
An encoded information sequence generating means for generating a code sequence group based on a predetermined code sequence, encoding each symbol forming the embedded information with the generated code sequence, and generating an encoded information sequence;
Two-dimensional information generating means for generating two-dimensional information arranged so that the encoded sequence information is an even function on a two-dimensional plane;
First transforming means for generating an embedded information image corresponding to the embedded information by performing an inverse Fourier transform on the two-dimensional information;
Information combining means for combining the embedded information image with the image of the document to be embedded;
An information embedding device comprising:   9. The information embedding apparatus according to claim 8, further comprising embedding information input means for inputting the embedding information and replacing the input embedding information with a preset symbol. Embedding information generated by performing inverse Fourier transform on two-dimensional information in which each symbol based on the embedding information is encoded with a predetermined code sequence group so that the encoded information becomes an even function on the two-dimensional plane. An information restoration apparatus for restoring the embedded information from a document or document data obtained by combining images,
Image reading means for reading image data from the document or document data;
Second transforming means for performing a Fourier transform on the image data to generate a frequency space representation of the image data;
Extraction information sequence generation means for generating an extraction information sequence by serially extracting real number components from the frequency space representation;
Embedded information restoring means for computing the correlation value of each code sequence of the extracted information sequence and the code sequence group, and restoring the embedded information by determining each symbol of the embedded information from the computation result;
An information restoration apparatus comprising:

Images