JP4172586B2 - Digital watermark insertion method, digital watermark insertion apparatus, digital watermark insertion program, digital watermark detection method, digital watermark detection apparatus, and digital watermark detection program - Google Patents

Description

  The present invention relates to a digital watermark insertion method for inserting a digital watermark into image data, its apparatus and its program, and a digital watermark detection method for detecting a digital watermark from image data into which the digital watermark has been inserted, its apparatus and its Regarding the program.

  In recent years, a technique for inserting a digital watermark into image data has been developed, and information for identifying a person who has a copyright, etc., information on usage methods and conditions for permitting the use of a copyrighted work, etc. in the form of a digital watermark. Insertion into image data has been put into practical use.

  Normally, a digital watermark is inserted and detected in the form of electronic data, but a digital watermark corresponding to printing has also appeared. This makes it possible to detect a digital watermark from the printed image data even when the image data into which the digital watermark is inserted is printed.

  For example, the scale is indefinite, it is rotated, the resolution is low, camera shake occurs, the signal-to-noise ratio is poor, and the embedded data in the form of a digital watermark is stabilized even for images with poor contrast. A robust digital watermark insertion method, apparatus and program thereof, and digital watermark detection method, apparatus and program thereof have been invented and disclosed in Patent Document 1. ing.

On the other hand, in recent years, game machines using cards for children are appearing in game centers and the like. Information is written in the form of a barcode on the card used in the game machine, and the barcode is read by the card reader of the game machine, and the game machine executes the game using the information.
JP 2004-15396 A

  Therefore, instead of writing information in the form of a barcode on a card used in a game machine, it has been proposed to write information on the card in the form of a digital watermark in accordance with the invention described in Patent Document 1. Then, you can write a lot of information on the card, and the appearance of the card doesn't deteriorate.

  However, in order to read the information written in the form of digital watermark on the card, it is necessary to read the image printed on the card with an image scanner, but if the read image is Fourier transformed, the frequency domain In particular, noise is generated. That is, many of the images of the image include lines and bands, but such lines and bands become frequency components aligned with radiation starting from the origin in the frequency domain. Further, the card is not always read in the correct direction, and may be read in a tilted state. In such a case, the contour of the card is read. The contour read in this way is also a frequency component aligned with radiation starting from the origin in the frequency domain. In addition, since the two sides of the card form 90 degrees, in the frequency domain, frequency components are generated in radiation having an origin of 90 degrees with respect to each other. Furthermore, although considered to be due to the characteristics of the discrete Fourier transform, frequency components are generated along the X axis and the Y axis in the frequency domain.

  If there are extra frequency components in this way, they become noise components, and a digital watermark that should originally be detected may not be detected. In other words, the digital watermark that should be detected may be buried in noise, or the noise may be erroneously detected as a digital watermark.

  Therefore, the present invention provides a digital watermark insertion method, a digital watermark insertion apparatus, and a digital watermark insertion that enable data embedded in the form of a digital watermark in a card read by an image scanner to be detected with high error tolerance. It is an object to provide a program, a digital watermark detection method, a digital watermark detection apparatus, and a digital watermark detection program.

According to a first aspect of the present invention, a step of generating a non-zero frequency component for detecting a rotation angle of an image in a frequency domain, and a first predetermined number of candidate positions in the frequency domain Generating a non-zero frequency component at a second predetermined number that is less than the first predetermined number and is interlaced with the data to be inserted, and combining the generated non-zero frequency component Converting the frequency domain to the spatial domain, and adding or subtracting the image obtained by the conversion to the original image . A value for each position is detected by taking a majority vote between each set of a predetermined number of candidate positions, and the value is obtained by taking the majority vote. By preparative sees whether the number of detected position is the second predetermined number, in order to the error detection, a plurality of said first predetermined number of sets of candidate positions, in one set There is provided a digital watermark insertion method characterized in that a candidate position of the same number as the certain number belonging to another set does not overlap on a straight line connecting a candidate position of a certain number and the origin.

  In the above-described electronic watermark insertion method, a plurality of sets of the first predetermined number of candidate positions are provided, and candidates for all numbers belonging to other groups on a straight line connecting a candidate position of a certain number belonging to a certain group and the origin The positions may not overlap.

  In the above digital watermark insertion method, the non-zero frequency component for detecting the rotation angle of the image includes a plurality of non-zero frequency components, and another non-zero frequency on a straight line connecting a certain non-zero frequency component and the origin. You may make it a component not overlap.

  In the above-described digital watermark insertion method, the non-zero frequency component for detecting the rotation angle of the image includes a plurality of non-zero frequency components, and the other is on a straight line orthogonal to a straight line connecting a certain non-zero frequency component and the origin. The non-zero frequency components may not overlap.

  In the above digital watermark insertion method, the non-zero frequency component for detecting the rotation angle of the image is a non-zero frequency component arranged in a circle with a first radius and a non-zero frequency component arranged in a circle with a second radius. And a zero frequency component.

According to a second aspect of the present invention, obtaining one or more image blocks from an image, converting data of the one or more image blocks from a spatial domain to a frequency domain, Obtaining an amplitude of each frequency component based on each frequency component; and detecting a non-zero frequency component for detecting a rotation angle of the image based on the amplitude of each frequency component in the frequency domain. Detecting the rotation angle of the image using the rotation angle, correcting the position of the amplitude of each frequency component in the frequency domain based on the rotation angle, and based on the amplitude of each frequency component subjected to the correction in the frequency domain. In addition, among the first predetermined number of candidate positions, positions that are combined with each other by data and having a second predetermined number smaller than the first predetermined number Detecting the data by detecting a non-zero frequency component that has been input, wherein a plurality of the first predetermined candidate position sets are provided, and the detection In the step, a value for each position is detected by taking a majority vote between each set of the first predetermined number of candidate positions, and the predetermined value is obtained by taking the majority vote. An electronic watermark detection method is provided , further comprising a step of detecting an error by checking whether or not the number of detected positions is the second predetermined number .

  In the above-described digital watermark detection method, a candidate position with the same number as the certain number belonging to another set may not overlap with a straight line connecting the candidate position with a certain number belonging to a certain set and the origin.

  In the above-described digital watermark detection method, all the number candidate positions belonging to another set may not overlap on a straight line connecting the candidate position of a number belonging to a set and the origin.

  In the above-described digital watermark detection method, the non-zero frequency component for detecting the rotation angle of the image is a non-zero frequency component arranged in the first radius circle and a non-zero frequency component arranged in the second radius circle. Detecting the rotation angle includes a predetermined number of non-zero frequency components and a circle of the second radius in order of amplitude among the frequency components in the circle of the first radius. The determination may be performed based on a predetermined number of non-zero frequency components in the order of amplitude among certain frequency components.

  In the above digital watermark detection method, other non-zero frequency components do not overlap on a straight line connecting a certain non-zero frequency component of the non-zero frequency components for detecting the rotation angle of the image and the origin, The non-zero frequency component does not overlap another straight line that is orthogonal to the straight line connecting the non-zero frequency component and the origin, and the step of detecting the rotation angle is performed for the origin among the upper predetermined number of non-zero frequency components. There may be provided a step of deleting what exists at the intersection of two straight lines orthogonal to each other and the circle of the first radius or the second circle.

  According to the present invention, since data is detected by majority vote, error tolerance is enhanced.

  Further, according to the present invention, since the total number is checked against the result of the majority decision, the error resistance is further enhanced.

  Furthermore, according to the present invention, since the arrangement of non-zero frequency components is determined in consideration of noise characteristics, error tolerance is enhanced.

  The best mode for carrying out the present invention will be described below in detail with reference to the drawings.

[pattern]
FIG. 1 is a diagram showing a pattern in the frequency domain of a digital watermark according to an embodiment of the present invention. FIG. 1 shows only the first quadrant and the second quadrant in the frequency domain, and omits the third and fourth quadrants. This is because the frequency component amplitude is a point about the origin of the frequency domain. This is because they are distributed symmetrically.

  As shown in FIG. 1, a rotation angle inner circle, a circle group 1 and a rotation angle outer circle exist in the frequency domain.

  A non-zero frequency component is arranged in the rotation angle inner circle and the rotation angle outer circle as shown by black circles in FIG. That is, non-zero frequency components C and D are arranged in the rotation angle inner circle, and non-zero frequency components A and B are arranged in the rotation angle outer circle. When detecting data from a digital watermark, the rotation angle of the image is detected based on the non-zero frequency components A, B, C, and D as preprocessing. The reason why the non-zero frequency components A, B, C, and D for detecting the rotation angle are arranged as shown in FIG. 2 will be described later.

As shown in FIG. 1, area 1 is arranged in circle group 1. In area 1, a non-zero frequency component or a zero frequency component is arranged as shown in FIG. Here, the fact that the zero frequency component is arranged means that a non-zero frequency component is not arranged. The circle group 1 includes an element circle 1, an element circle 2, and an element circle 3. As shown in FIG. 3, non-zero frequency components are arranged at five positions out of ten positions 1-1 to 1-10 of the element circle 1, and zero frequency components are arranged at the remaining five positions. Is placed. The five positions where the non-zero frequency components are arranged are determined according to the value of data to be embedded. Therefore, ₁₀ C ₅ = 252 types of data can be embedded. Similarly, non-zero frequency components are arranged at five positions out of ten positions 1-1 to 1-10 of the element circle 2, and zero frequency components are arranged at the remaining five positions, Non-zero frequency components are arranged at five positions among the ten positions 1-1 to 1-10 of the element circle 3, and zero frequency components are arranged at the remaining five positions.

  The same data is arranged at the positions assigned the same numbers between the element circles 1 to 3. That is, for example, when a non-zero frequency component is arranged at the position 1-1 of the element circle 1, the non-zero frequency component is arranged also at the position 1-1 of the element circle 2 and the position 1-1 of the element circle 3. When the zero frequency component is arranged at the position 1-1 of the element circle 1, the zero frequency component is also arranged at the position 1-1 of the element circle 2 and the position 1-1 of the element circle 3. The same applies to other positions.

  Moreover, the phase seen in the polar coordinate of the frequency domain of the position of the same number differs for every element circle. Therefore, the position of the number of another element circle does not exist on the straight line connecting the position of the number of a certain element circle and the origin.

  Furthermore, no matter how the combination of numbers is selected, there is no position of a certain number of another element circle on a straight line connecting the position of the certain number of a certain element circle and the origin. That is, two or more non-zero frequency components or zero frequency components do not exist on the radiation extending from the origin in the frequency domain.

  Furthermore, the phase difference in the frequency domain between the element circles at the same number position is about 60 degrees, which is a value obtained by dividing about 180 degrees into three equal parts. That is, the phases in the frequency region at the same number positions are distributed evenly. This is to make the interleave length as long as possible.

  By selecting the position where the non-zero frequency component or the zero frequency component is arranged in this way, it is possible to obtain a high error correction capability and a high error detection capability for burst errors along the radiation extending from the origin in the frequency domain. In many cases, lines or bands are included in the content of the original image, and in such a case, many non-zero frequency components are generated along the radiation extending from the origin in the frequency domain. In addition, when the card is read in an inclined state, the outline of the card is read obliquely, and even in such a case, many non-zero frequency components are generated along the radiation extending from the origin in the frequency domain. . Therefore, it is effective to improve the error correction capability and error detection capability for burst errors along the radiation extending from the origin in the frequency domain.

FIG. 4 shows a pattern in which the number of areas is increased compared to the pattern shown in FIG. In the pattern shown in FIG. 1, the number of areas is one, and only ₁₀ C ₅ = 252 data can be represented. However, in the pattern shown in FIG. 4, the number of areas is six and ( ₁₀ C ₅ ) ^ 6 = 252 ^ 6 = 2.561 × 10 ^ 14 kinds of data can be represented. FIG. 4 schematically shows the arrangement of the areas for explanation, and the actual area shape is different. For example, area 2 and area 3 are mixed in circle group 2, and positions 2-1 to 2-10 included in area 2 and positions 3-1 to 3-10 included in area 3 are as shown in FIG. Be placed. This is to make the interleave length as long as possible.

[Angle detection method]
Next, a method of detecting an image rotation angle using a pattern as shown in FIG. 2 will be described with reference to FIG.

  Referring to FIG. 6, first, four amplitude points are selected in descending order of amplitude from the outer circle of the rotation angle (step S301). Next, four amplitudes are selected in descending order of amplitude from the rotation angle inner circle (step S303). Next, step S307 is repeated for the four points of the rotation angle outer circle detected in step S301 (step S305). In step S307, when the point of interest is point A or point B (see FIG. 2), three or four of the points selected in steps S301 and S303 are rotation angle detection patterns. It is determined whether or not three or four of the four points A, B, C and D coincide with each other. If so (YES in step S307), the rotation angle is detected from the three or four points (step S309).

  If YES in step S307 during the repetition of step S305, the process proceeds to step S311 to mark a non-processing point to be excluded from the processing target. That is, the fact that the answer in step S307 was not YES is that the point selected in step S301, step S303 or both may include a point that is noise. The noise is likely to be noise distributed on two axes orthogonal to each other. Therefore, among the points selected in step S301, step S303 or both, three or four points forming one of the five types of patterns as shown in FIG. Mark as external point).

  Next, except for the marked points, four amplitude points are selected in descending order of amplitude from the rotation angle outer circle (step S313). Next, except for the marked points, four amplitudes are selected in descending order of amplitude from the rotation angle inner circle (step S315). Next, step S3219 is repeated for the four points of the rotation angle outer circle detected in step S313 (step S317). In step S319, when the point of interest is point A or point B (see FIG. 2), three or four of the points selected in steps S313 and S315 are rotation angle detection patterns. It is determined whether or not three or four of the four points A, B, C and D coincide with each other. If so (YES in step S319), the rotation angle is detected from the three or four points (step S321).

  If YES in step S319 during the repetition of step S317, the process ends in error.

[Data detection method]
Next, a method for detecting data from a pattern as shown in FIG. 3 will be described with reference to FIG.

  First, five non-zero frequency components are detected from ten positions where non-zero frequency components or zero frequency components are arranged for each element circle in each area (steps S351 and S353). In order to detect five non-zero frequency components, after correcting the position of the frequency component amplitude in the frequency domain by the detected rotation angle, the non-zero frequency component or the zero frequency component is arranged for the amplitude of each frequency component. Through the window that passes only the 10 positions or their amplitudes in the vicinity of the position, and then gradually lowers the threshold value from the maximum amplitude, and 5 frequency components having amplitudes above the threshold value. When detected, the five frequency components are assumed to be non-zero frequency components (see FIG. 9). The value of the non-zero frequency component is set to 1, and the value of the zero frequency component is set to 0.

  Next, a majority vote is taken between element circles for each position in the area (steps S357 and S359). That is, for example, a majority decision is made among the value of the position 1-1 of the element circle 1, the value of the position 1-1 of the element circle 2, and the value of the position 1-1 of the element circle 3. By performing this operation for 10 positions, 10-bit data in the area can be obtained.

  Next, error detection is performed (step S363). That is, it is determined whether or not the value of 5 bits out of 10 bits in the area is 1 and the value of the remaining 5 bits is 0. If so (YES in step S363), it is assumed that the values of all bits (10 bits) obtained by the majority decision are correct (step S365). Otherwise, it is determined that there is an erroneous bit among all the bits (10 bits) obtained by the majority decision, and data detection is impossible (error detection) (step S367).

  FIG. 10 shows an example when data detection is possible. In the example shown in FIG. 10, two bits of one element circle are incorrect, but a correct value is obtained by majority vote, and no error is detected even by error detection.

  FIG. 11 shows an example when data detection is impossible. In the example shown in FIG. 11, no error occurs in one element circle, but an error of 2 bits occurs in the remaining two element circles. As a result of the majority decision, two bits can be detected correctly, but one bit is detected incorrectly. When an error is detected, the number of 1s is 6 instead of 5, so that data cannot be detected.

  Similarly, in the case of the pattern shown in FIG. 4, data is detected for each area. At this time, the detection of the five non-zero frequency components is performed based on the frequency components at ten positions of each element circle in each area.

[Digital watermark insertion device]
FIG. 12 is a block diagram showing the configuration of the digital watermark insertion apparatus according to the embodiment of the present invention.

  Referring to FIG. 12, a digital watermark insertion apparatus according to an embodiment of the present invention includes a rotation angle detection pattern generation unit 101, a data holding unit 102, a data pattern generation unit 103, a synthesis unit 104, an inverse discrete Fourier transform unit 105, A valuation unit 106 and an addition unit 107 are provided. These parts can be realized by hardware, but can also be realized by reading and executing a program for causing a computer to function as these parts.

  Next, a digital watermark insertion method performed by the digital watermark insertion apparatus will be described with reference to FIGS.

  The rotation angle detection pattern generation unit 101 generates a rotation angle detection pattern as shown in FIG. 2 (step S151). The data holding unit 102 holds data to be inserted. The data pattern generation unit 103 generates a data pattern as shown in FIG. 3 or 4 in accordance with the data value held in the data holding unit 102 (step S152). The combining unit 104 combines the rotation angle detection pattern generated by the rotation angle detection pattern generation unit 101 and the data pattern generated by the data pattern generation unit 103 to generate a pattern as shown in FIG. 1 (step S153). ). The inverse Fourier transform unit 105 performs inverse Fourier transform on the pattern as shown in FIG. 1 generated by the synthesis unit 104 to generate a pattern in the spatial domain (spatial domain pattern) (step S154). The binarization unit 106 binarizes the space area pattern generated by the inverse Fourier transform unit 105 (step S155). The adding unit 107 adds the spatial region pattern binarized by the binarizing unit 106 to all or some of the yellow components of the original image (step S156).

  Note that a spatial region pattern that does not pass through the binarization unit 106 may be added to the yellow component of the original image. Alternatively, the adding unit 107 may be replaced with a subtracting unit, and a binarized spatial region pattern or a non-binarized spatial region pattern may be subtracted from the yellow component of the original image.

[Digital watermark detection device]
FIG. 14 is a block diagram showing the configuration of the digital watermark detection apparatus according to the embodiment of the present invention.

  Referring to FIG. 14, a digital watermark detection apparatus according to an embodiment of the present invention includes a spatial filter 201, a blocking unit 202, a discrete Fourier transform unit 203, an amplitude calculation unit 204, an amplitude integration unit 205, a rotation angle detection unit 206, and data. A detection unit 207 is provided. These parts can be realized by hardware, but can also be realized by reading and executing a program for causing a computer to function as these parts.

  Next, a digital watermark detection method performed by the digital watermark detection apparatus shown in FIG. 14 will be described with reference to FIGS.

  As is clear from the fact that the spatial filter 201 has a weight as shown in FIG. 16, it has a band pass characteristic. The spatial filter 201 is applied to the blue component (digital watermark inserted image) of the image into which the digital watermark has been inserted (step S251). Thereby, non-zero frequency components other than the pattern shown in FIG. 1 are suppressed. The blocking unit 202 blocks the digital watermark insertion image to which the spatial filter is applied into one or more blocks (step S252). The discrete Fourier transform unit 203 performs discrete Fourier transform on the digital watermark insertion image of each block output from the blocking unit 202 to obtain each frequency component of each block (step S253). The amplitude calculator 204 calculates the amplitude of each frequency component of each block (step S254). The amplitude integrating unit 205 integrates the amplitude for each frequency between the blocks (step S255). The rotation angle detection unit 206 detects the rotation angle of the image based on the amplitude of each frequency component output from the amplitude integration unit 205 (step S256). The method for detecting this rotation angle is as described above. The data detection unit 208 detects data based on the amplitude output from the amplitude integration unit 205 and the rotation angle output from the rotation angle detection unit 206 (step S257). The method for detecting this data is as described above.

[Angle detection pattern]
As described above, the angle detection pattern is as shown in FIG. 2, and the reason why the angle detection pattern is used will be described below.

  The angle detection pattern satisfies the following conditions.

  Condition 1: Two or more points constituting the angle detection pattern do not exist on the same radiation.

  Condition 2: Two or more points constituting the angle detection pattern are not in an angular relationship of 90 degrees with respect to the origin.

  Condition 3: The points constituting the angle detection pattern are separated from the X axis and the Y axis.

  Condition 4: The rotation angle can be detected based on three of the four points constituting the angle detection pattern.

  As described above, the condition 1 often includes a line or a band in the contents of the original image. In such a case, many non-zero frequency components along the radiation extending from the origin are generated in the frequency domain. In addition, when the card is read in an inclined state, the outline of the card is read obliquely, and even in such a case, many non-zero frequency components are generated along the radiation extending from the origin in the frequency domain. In consideration of the phenomenon that the angle detection pattern is formed, the point that does not constitute the angle detection pattern is provided in order to prevent erroneous recognition as the angle detection pattern. In addition, if two or more points exist in the same radial shape, when the two or more points overlap with the X axis or Y axis, the two or more points can be distinguished from noise on the X axis or Y axis. The rotation angle must be detected from the remaining two or less points. However, since this is difficult, Condition 1 is provided.

  Condition 2 is that when the card is read in a tilted state, the contours of the two sides of the card are read diagonally, and in the frequency domain, a non-zero frequency is placed on two radiations that are at an angular relationship of 90 degrees with respect to the origin. Considering the phenomenon that a large number of components occur, it is provided in order to prevent the point that does not constitute the angle detection pattern from being erroneously recognized as the angle detection pattern. Also, if two or more points form a 90-degree angular relationship with respect to the origin, the two or more points may overlap the X and Y axes at the same time due to the rotation of the card. The two or more points cannot be distinguished from the noise on the X-axis or Y-axis, and the rotation angle must be detected from the remaining two or less points. 2 was provided.

  Condition 3 considers the phenomenon that many non-zero frequency components are generated on the X-axis and the Y-axis and that the card is read without tilting so that the angle detection pattern does not constitute an angle detection pattern. It was set up to prevent it from being mistakenly recognized.

  Condition 4 is that if the angle detection pattern is composed of two points, it is difficult to detect the points constituting the angle detection pattern separately from the points not constituting the angle detection pattern. Assuming that the pattern for use is composed of three points, the difficulty is alleviated, but since it is not always possible to detect the three points, four points are preferably provided redundantly. If four points are provided, the rotation angle can be detected from the remaining three points even when one point cannot be detected by overlapping the X axis or the Y axis.

  The reason why the points constituting the angle detection pattern are arranged on the outermost circle is that the longer the radius of the circle, the higher the angle detection accuracy.

  Furthermore, the reason why the points constituting the angle detection pattern are arranged in the innermost circle is as follows. That is, the innermost circle cannot be used as a circle into which data is inserted because there is a lot of noise due to the original image in the periphery and the amount of data that can be embedded is small. Therefore, if data is inserted into a circle having a larger radius than the innermost circle, the innermost circle remains. When the angle detection pattern is inserted into the innermost circle, the above-described problem hardly occurs.

[Error correction probability]
The probability of error correction indicates that an error occurs at two points out of three points representing a certain bit (first bit) of the same group, and further represents the other bit (second bit) of the group. It can be approximated by the probability of occurrence in the main case, which is the case where an error occurs at two of the three points. Then, the first bit is inverted and the second bit is also inverted, so that the sum is not changed, and the error cannot be corrected by the sum check. Furthermore, the probability when more errors occur can be ignored.

  Therefore, the probability P when erroneous correction occurs is calculated next.

  If the probability that an error occurs at one point is Pa, the probability that an error occurs at two points is Pa ^ 2. This probability is according to a method of detecting five non-zero frequencies from ten candidate positions, and is lower than the probability that an error occurs independently at each point.

  The probability Pb in the case where these two errors overlap in terms of representing the same bit in the same group is equal to the probability Pc in the case where these two errors are in the same group. This is the probability Pd in the case of overlapping at the point representing the same bit in the same group (Pb = Pc × Pd).

  In the example of FIG. 4, the probability Pc indicates that an error occurs at any point in the second circle of a group when the first error occurs at a point in the first circle of the group. 1 / (number of element circles in area × number of areas) = 1 / (3 × 6), and the probability that an error will occur at any point in the third circle of the group It is 2/18 obtained by adding a certain 1 / (number of element circles in area × number of areas) 1 / (3 × 6).

  The probability Pd is calculated when one of 10 points of the same radius of the same group is inverted from 1 to 0 and the other one of the 10 points is inverted from 0 to 1. The probability is 1 / (5 × 5) = 1/25. This probability includes the probability of occurrence of two errors in addition to the two errors of interest.

Therefore, the probability P in the case of erroneous correction occurs is
P = Pa ^ 2 × Pb
= Pa ^ 2 * Pc * Pd
It becomes. Substituting the experimental value Pa = 1/48,
P = 1/48 x 1/48 x 2/18 x 1/25
= 1 / 518,400
It becomes.

[Application example]
The copy prohibition code is printed on the printed matter in the form of a digital watermark according to the present invention. The copier is provided with the digital watermark detection apparatus and the copy prohibition unit according to the above embodiment. When the scanner unit of the copier reads the printed matter on which the copy prohibition code is printed, the digital watermark detection device detects the copy prohibition code from the read matter and the copy prohibition unit operates the printing unit of the copier. To stop.

  The digital watermark of the above embodiment is printed on a printed matter with ink that reflects or radiates only light of a wavelength (wavelength other than visible light wavelength) that cannot be copied by a normal copier. In this case, the digital watermark is printed directly on such special ink without passing through the adder 107. Then, even if this printed matter is copied with a normal copying machine, the digital watermark portion cannot be copied. Then, if an apparatus that performs a security check by detecting a digital watermark printed with special ink, the authenticity of the printed matter can be determined. As such printed matter, for example, a license, an identification card, a passport, an employee ID card, etc. can be assumed.

  The present invention can be used to print data on a card so that the data is not visually noticeable.

It is a figure which shows the pattern in the frequency domain of the digital watermark by embodiment of this invention. It is a figure which shows the pattern for rotation angle detection among the patterns shown in FIG. It is a figure which shows the pattern for data among the patterns shown in FIG. It is a figure which shows the pattern in the frequency domain of the digital watermark by other embodiment of this invention. It is a figure which shows a part of pattern for data in the pattern shown in FIG. 3 is a flowchart illustrating a rotation angle detection method according to an embodiment of the present invention. In the rotation angle detection method by embodiment of this invention, it is a figure showing the point deleted in a frequency domain. 3 is a flowchart illustrating a data detection method according to an embodiment of the present invention. It is a figure for demonstrating the data detection method by embodiment of this invention. It is a 1st figure for demonstrating the error detection method by embodiment of this invention. It is a 2nd figure for demonstrating the error detection method by embodiment of this invention. It is a block diagram which shows the structure of the digital watermark insertion apparatus by embodiment of this invention. 5 is a flowchart illustrating a digital watermark insertion method according to an embodiment of the present invention. It is a block diagram which shows the structure of the digital watermark detection apparatus by embodiment of this invention. 3 is a flowchart illustrating a digital watermark detection method according to an embodiment of the present invention. It is a figure which shows the tap coefficient of the spatial filter shown in FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 Rotation angle detection pattern generation part 102 Data holding part 103 Data pattern generation part 104 Synthesis | combination part 105 Inverse discrete Fourier transform part 106 Binarization part 107 Addition part (subtraction part)
201 Spatial Filter 202 Blocking Unit 203 Discrete Fourier Transform Unit 204 Amplitude Calculation Unit 205 Amplitude Integration Unit 206 Rotation Angle Detection Unit 207 Data Detection Unit

Claims (23)

Generating a non-zero frequency component for detecting a rotation angle of the image in the frequency domain;
In the frequency domain, a non-zero frequency component is added to a second predetermined number of positions that are mutually combined by data to be inserted among the first predetermined number of candidate positions and that is less than the first predetermined number. Generating step;
Combining the generated non-zero frequency components together from the frequency domain to the spatial domain;
Adding or subtracting the image obtained by the conversion to the original image;
In a digital watermark insertion method comprising:
In the electronic watermark detection method, a value for each position is detected by taking a majority vote between each set of the first predetermined number of candidate positions, and the predetermined value is obtained by taking the majority vote. In order to detect an error by checking whether the number of detected positions is the second predetermined number, a plurality of sets of the first predetermined number of candidate positions are provided,
A digital watermark insertion method, wherein a candidate position of the same number as the certain number belonging to another set does not overlap a straight line connecting a candidate position of a certain number belonging to a certain set and the origin. The digital watermark insertion method according to claim 1,
A plurality of sets of the first predetermined number of candidate positions are provided so that candidate positions of all numbers belonging to other sets do not overlap on a straight line connecting a candidate position of a number belonging to a set and the origin. A method for inserting a digital watermark. The digital watermark insertion method according to claim 1,
The non-zero frequency component for detecting the rotation angle of the image includes a plurality of non-zero frequency components, and other non-zero frequency components do not overlap on a straight line connecting a certain non-zero frequency component and the origin. To insert digital watermark. The digital watermark insertion method according to claim 1,
The non-zero frequency component for detecting the rotation angle of the image includes a plurality of non-zero frequency components, and other non-zero frequency components do not overlap with a straight line orthogonal to a straight line connecting a certain non-zero frequency component and the origin. A digital watermark insertion method characterized by the above. The digital watermark insertion method according to claim 1,
The non-zero frequency component for detecting the rotation angle of the image includes a non-zero frequency component arranged in a circle with a first radius and a non-zero frequency component arranged in a circle with a second radius. A digital watermark insertion method as a feature. Obtaining one or more image blocks from the image;
Converting the data of the one or more image blocks from a spatial domain to a frequency domain;
Obtaining the amplitude of each frequency component based on each frequency component of the one or more image blocks;
Detecting the rotation angle of the image in the frequency domain by detecting a non-zero frequency component for detecting the rotation angle of the image based on the amplitude of each frequency component;
Correcting the position in the frequency domain of the amplitude of each frequency component by the rotation angle;
In the frequency domain, based on the amplitude of each frequency component subjected to the correction, the second predetermined position is a position that is mutually combined by data and is smaller than the first predetermined number. Detecting the data by detecting non-zero frequency components inserted into a predetermined number of
In a digital watermark detection method comprising:
A plurality of sets of the first predetermined candidate positions are provided;
In the detection step, a value for each position is detected by taking a majority vote between each set of the first predetermined number of candidate positions ,
A digital watermark further comprising a step of detecting an error by checking whether or not the number of positions detected to be a predetermined value by taking the majority decision is the second predetermined number. Detection method. The digital watermark detection method according to claim 6.
A digital watermark detection method, wherein a candidate position of the same number as the certain number belonging to another set does not overlap on a straight line connecting a candidate position of a certain number belonging to a certain set and the origin. The digital watermark detection method according to claim 7 ,
A digital watermark detection method, wherein candidate positions of all numbers belonging to another set do not overlap on a straight line connecting a candidate position of a number belonging to a set and the origin. The digital watermark detection method according to claim 6.
The non-zero frequency component for detecting the rotation angle of the image includes a non-zero frequency component arranged in a first radius circle and a non-zero frequency component arranged in a second radius circle,
The step of detecting the rotation angle includes: an upper predetermined number of non-zero frequency components in order of amplitude among frequency components in the first radius circle and an amplitude of frequency components in the second radius circle. An electronic watermark detection method, which is performed based on a predetermined number of non-zero frequency components in order. The digital watermark detection method according to claim 9 .
Other non-zero frequency components do not overlap on a straight line connecting a certain non-zero frequency component of the non-zero frequency components for detecting the rotation angle of the image and the origin, and a certain non-zero frequency component and the origin are Other non-zero frequency components do not overlap on the straight line orthogonal to the connecting line,
The step of detecting the rotation angle is present at the intersection of two straight lines orthogonal to each other with respect to the origin and the circle of the first radius or the second circle of the upper predetermined number of non-zero frequency components. A method for detecting a digital watermark, comprising: deleting a thing. Means for generating a non-zero frequency component for detecting the rotation angle of the image in the frequency domain;
In the frequency domain, a non-zero frequency component is added to a second predetermined number of positions that are mutually combined by data to be inserted among the first predetermined number of candidate positions and that is less than the first predetermined number. Means for generating;
Means for combining the generated non-zero frequency components into a frequency domain to a spatial domain;
Means for adding or subtracting the image obtained by the conversion to the original image;
In the digital watermark insertion device, comprising:
In the digital watermark detection apparatus, a value for each position is detected by taking a majority vote between each set of the first predetermined number of candidate positions, and the predetermined value is obtained by taking the majority vote. In order to detect an error by checking whether the number of detected positions is the second predetermined number, a plurality of sets of the first predetermined number of candidate positions are provided. An electronic watermark inserting apparatus characterized in that a candidate position of the same number as the certain number belonging to another set does not overlap a straight line connecting a candidate position of a certain number and the origin. The digital watermark insertion apparatus according to claim 11 , wherein
A plurality of sets of the first predetermined number of candidate positions are provided so that candidate positions of all numbers belonging to other sets do not overlap on a straight line connecting a candidate position of a number belonging to a set and the origin. An electronic watermark insertion apparatus characterized by the above. The digital watermark insertion apparatus according to claim 11 , wherein
The non-zero frequency component for detecting the rotation angle of the image includes a plurality of non-zero frequency components, and other non-zero frequency components do not overlap on a straight line connecting a certain non-zero frequency component and the origin. Digital watermark insertion device. The digital watermark insertion apparatus according to claim 11 , wherein
The non-zero frequency component for detecting the rotation angle of the image includes a plurality of non-zero frequency components, and other non-zero frequency components do not overlap with a straight line orthogonal to a straight line connecting a certain non-zero frequency component and the origin. A digital watermark insertion apparatus characterized by the above. The digital watermark insertion apparatus according to claim 11 , wherein
The non-zero frequency component for detecting the rotation angle of the image includes a non-zero frequency component arranged in a circle with a first radius and a non-zero frequency component arranged in a circle with a second radius. A digital watermark insertion device. Means for obtaining one or more image blocks from the image;
Means for transforming the data of the one or more image blocks from a spatial domain to a frequency domain;
Means for obtaining the amplitude of each frequency component based on each frequency component of the one or more image blocks;
Means for detecting the rotation angle of the image in the frequency domain by detecting a non-zero frequency component for detecting the rotation angle of the image based on the amplitude of each frequency component;
Means for correcting the position of the amplitude of each frequency component in the frequency domain by the rotation angle;
In the frequency domain, based on the amplitude of each frequency component subjected to the correction, the second predetermined position is a position that is mutually combined by data and is smaller than the first predetermined number. Means for detecting said data by detecting non-zero frequency components inserted into a predetermined number of
In an electronic watermark detection apparatus comprising:
A plurality of sets of the first predetermined candidate positions are provided;
In the detection means, a value for each position is detected by taking a majority vote between each set of the first predetermined number of candidate positions .
A digital watermark , further comprising means for detecting an error by checking whether or not the number of positions detected to be a predetermined value by taking the majority decision is the second predetermined number. Detection device. The digital watermark detection apparatus according to claim 16 , wherein
An electronic watermark detection apparatus, wherein a candidate position of the same number as the certain number belonging to another set does not overlap a straight line connecting a candidate position of a certain number belonging to a certain set and the origin. The digital watermark detection apparatus according to claim 17 ,
An electronic watermark detection apparatus characterized in that candidate positions of all numbers belonging to another set do not overlap on a straight line connecting a candidate position of a number belonging to a set and the origin. The digital watermark detection apparatus according to claim 17,
The non-zero frequency component for detecting the rotation angle of the image includes a non-zero frequency component arranged in a first radius circle and a non-zero frequency component arranged in a second radius circle,
The means for detecting the rotation angle includes an upper predetermined number of non-zero frequency components in order of amplitude of frequency components in the circle of the first radius and an amplitude of frequency components in the circle of the second radius. An electronic watermark detection apparatus, which performs the processing based on a predetermined number of non-zero frequency components in order. The digital watermark detection apparatus according to claim 19 , wherein
Other non-zero frequency components do not overlap on a straight line connecting a certain non-zero frequency component of the non-zero frequency components for detecting the rotation angle of the image and the origin, and a certain non-zero frequency component and the origin are Other non-zero frequency components do not overlap on the straight line orthogonal to the connecting line,
The means for detecting the rotation angle exists at the intersection of two straight lines orthogonal to each other with respect to the origin and the circle of the first radius or the second circle of the upper predetermined number of non-zero frequency components. An electronic watermark detection apparatus comprising means for deleting an object. The digital watermark detection apparatus according to claim 16 or 20 ,
Means for prohibiting copying of the image when predetermined data is detected by the digital watermark detection device;
A copying apparatus comprising:   An electronic watermark insertion program for causing a computer to perform the electronic watermark insertion method according to any one of claims 1 to 5. Digital watermark detection program for causing a digital watermark detection method according to any one of claims 6 to 10 in a computer.

Images