JP4541213B2 - Digital watermark insertion method, digital watermark detection method, digital watermark insertion device, and digital watermark detection device - Google Patents

Description

  The present invention relates to a digital watermark insertion method for inserting a digital watermark into an image, a digital watermark detection method for detecting a digital watermark from an image into which the digital watermark has been inserted, a digital watermark insertion apparatus and a digital watermark detection corresponding thereto. Relates to the device.

  A method of displaying a barcode on a printed matter or a homepage, photographing the barcode with a camera, and obtaining data such as a URL based on the photographed barcode is widely used. The code will reduce the appearance of printed materials and websites.

  In order to solve such a problem, a method of inserting a digital watermark representing data into an image has been proposed. If this method is used, data can be obtained from the captured image, so that a barcode is not necessary.

As an invention relating to such a method, the insertion apparatus sets a data pattern at a position to be combined by insertion data among an enlargement / reduction ratio detection pattern, a rotation angle detection pattern, and candidate positions, and reverses all patterns together. The Fourier transform is performed and the image obtained by the transform is added to the original image, and the detection device divides the image into blocks, Fourier transforms the data of each block to obtain the amplitude of each frequency component of each block, and each frequency The amplitude of the component is added between blocks to obtain the total amplitude for each frequency, the enlargement / reduction ratio of the image is detected with the enlargement / reduction ratio detection pattern, the rotation angle of the image is detected with the rotation angle detection pattern, The data is detected based on a combination of positions of a predetermined number of summed amplitudes indicating relatively large values among the summed amplitudes corrected for the enlargement / reduction ratio and the rotation angle at the candidate positions. Invention is (for example, see Patent Document 1). The present invention has a problem that the scale of the image is changed and the image is rotated especially when an image is taken with a camera, and thus it is impossible to detect a digital watermark representing data. In order to solve this problem, an electronic watermark for detecting the enlargement / reduction ratio and rotation angle of the image is embedded in the image.
JP 2004-15396 A

However, according to the invention described in the above-mentioned Patent Document 1, as shown in FIG. 1, in the frequency domain, in addition to the circle 101 where the non-zero frequency component representing the data is to be arranged, the enlargement / reduction ratio of the image is set. It is necessary to provide a circle 102 where a non-zero frequency component for detection is to be arranged and a circle 103 where a non-zero frequency component for detecting a rotation angle of an image is to be arranged. As a result, the number of circles where non-zero frequency components representing data are to be arranged is limited, and therefore the number of values that the data can take is limited. For example, in the example shown in FIG. 1, non-zero frequency components are arranged at 10 candidate positions out of 20 candidate positions on the circle 101, and the frequency components are point-symmetric with respect to the origin. Since non-zero frequency components are arranged at five candidate positions out of the ten candidate positions, the number of values that can be taken by the data is ₁₀ C ₅ = 252.

  Accordingly, an object of the present invention is to provide a digital watermark insertion method, a digital watermark detection method, a digital watermark insertion device, and a digital watermark detection device that can increase the number of values that can be taken by data.

According to the first aspect of the present invention, the non-zero frequency component generator generates a plurality of non-zero frequency components arranged on a straight line in the frequency domain viewed in polar coordinates (hereinafter referred to as “ non-zero frequency component ”). And a step of embedding the non-zero frequency component in the original image by embedding means .

In the above-described digital watermark insertion method, the embedding step includes a step of transforming the non-zero frequency component from the frequency domain to the spatial domain by a frequency domain / space domain transforming unit, and a step of embedding in the spatial domain by the spatial unit. Embedding the non-zero frequency component converted into a region in the original image.

In the above digital watermark insertion method, the data insertion unit may further include a data insertion step of embedding a digital watermark representing data in the original image.

According to the second aspect of the present invention, the step of converting the image from the spatial domain to the frequency domain by the spatial domain / frequency domain conversion means, and the frequency domain viewed in polar coordinates from the image converted to the frequency domain. A detection step of detecting a plurality of non-zero frequency components arranged on a straight line (hereinafter referred to as “ non-zero frequency component ”) by a non-zero frequency component detecting unit, and an enlargement / reduction ratio / rotation angle detecting unit, And a detection step of detecting an enlargement / reduction ratio and a rotation angle of the image based on the digital watermark detection method.

In the above-described digital watermark detection method, the detection step includes a value of a when the rotation angle is represented by θ and the radius is represented by r in the frequency domain viewed in polar coordinates, and the straight line is represented by θ = ar + b. And a step of detecting an image enlargement / reduction ratio, and a step of detecting a rotation angle of the image based on a value of b when the straight line is expressed as θ = ar + b. Also good.

An electronic watermark detecting method described above, by correcting means, and a correction step of correcting at the detected said scaling factor and rotation angle are the image, the data detection means, detecting a digital watermark representing the data from the corrected the image And a data detecting step.

  According to the present invention, the circle in the frequency domain is not used for embedding the pattern for detecting the enlargement / reduction ratio and the rotation angle, and therefore, the number of circles in which data can be embedded increases. You can increase the number.

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

  FIG. 2 shows a pattern in the frequency domain for detecting the enlargement / reduction ratio and rotation angle of an image according to an embodiment of the present invention. Reference numerals 201-1 to 201-11 and 201 ′-1 to 201 ′ -11 denote non-zero frequency components constituting the pattern. Since the frequency component is targeted at the origin, 201-1 to 201-11 correspond to 201'-1 to 201'-11, respectively.

  FIG. 3 shows the pattern shown in FIG. 2 when viewed in polar coordinates. The non-zero frequency components 201 "-1 to 201" -11 correspond to 201-1 to 201-11, respectively, and also correspond to 201'-1 to 201'-11. Non-zero frequency components 201 "-1 to 201" -11 are provided on a straight line 202 ". Curve 202 and curve 202 'in FIG. 2 correspond to straight line 202' 'in FIG. The straight line 202 ″ is represented by θ = a × r + b. In the example of FIGS. 2 and 3, b = π / 4 as apparent from the fact that the curve 202 and the curve 202 ′ are in contact with the straight line 203 having an inclination of π / 4 in FIG. 2.

  FIG. 4 shows how the straight line 202 ″ changes when the captured image is enlarged or reduced. As shown in FIG. 4, when the captured image is enlarged, the straight line 202 "changes to a straight line 202" A. That is, when the captured image is enlarged, the slope of the straight line increases. On the other hand, when the captured image is reduced, the straight line 202 "changes to a straight line 202" B. That is, when the captured image is reduced, the slope of the straight line is reduced. The slope of the straight line is represented by z × a, where z is the enlargement / reduction ratio.

  FIG. 5 shows how the straight line 202 ″ changes when the captured image is rotated. As shown in FIG. 5, when the captured image is rotated to the left, the straight line 202 "changes to a straight line 202" C. That is, when the photographed image is rotated to the left, the straight θ intercept becomes large. On the other hand, when the captured image is looking to the right, the straight line 202 "changes to a straight line 202" D. That is, when the image is rotated to the right, the θ intercept of the straight line becomes small. The θ intercept of the straight line is represented by b + Δb where the rotation angle is Δb.

  When the photographed image is enlarged or reduced and rotated, the slope of the straight line is expressed as z × a where the enlargement / reduction ratio is z, and the θ intercept of the straight line is If the rotation angle is Δb, it is expressed as b + Δb. Therefore, the straight line is represented by θ = (z × a) × r + (b + Δb). Since z × a and b + Δb are obtained from the detected straight lines, and a and b are known, z and Δb are obtained from these.

FIG. 6 shows a pattern in the frequency domain representing data. The circle 211 is provided with 20 candidate positions, and non-zero frequency components are arranged at 10 candidate positions according to the data value. In addition, since the distribution of frequency components is targeted at the origin, this arrangement is equivalent to the arrangement of non-zero frequency components according to data values at five candidate positions out of ten candidate positions. . Similarly, 28 candidate positions are provided in the circle 212, and non-zero frequency components are arranged in 14 candidate positions according to the data value. In addition, since the distribution of frequency components is targeted at the origin, this arrangement is equivalent to the arrangement of non-zero frequency components according to the data value at 7 candidate positions out of 14 candidate positions. . Therefore, according to this pattern, ₁₀ C ₅ × ₁₄ C ₇ = 864864 data can be represented. The circle 212 can be used to represent data by making the pattern for representing the enlargement / reduction ratio as shown in FIG. In the conventional example, the circle 212 has to be used for the pattern for detecting the enlargement / reduction ratio. Therefore, according to the present embodiment, it is possible to increase the number of values that data can take.

  The pattern shown in FIG. 7 is obtained by combining the pattern shown in FIG. 2 with the pattern shown in FIG. The pattern shown in FIG. 7 is subjected to inverse Fourier transform and superimposed (added or subtracted) on the original image.

  Next, a digital watermark insertion apparatus according to an embodiment of the present invention will be described.

  Referring to FIG. 8, the digital watermark insertion apparatus according to the present embodiment includes a correction pattern generation unit 301, a data holding unit 302, a data pattern generation unit 303, a synthesis unit 304, an inverse discrete Fourier transform unit 305, and a binarization unit 306. And an addition unit (or subtraction unit) 307.

  The correction pattern generation unit 301 is a digital watermark detection apparatus that generates a pattern (such as shown in FIG. 2) in the frequency domain for detecting the enlargement / reduction ratio and rotation angle of the input image.

  The data holding unit 302 holds data to be embedded in the image in the form of a digital watermark.

  The data pattern generation unit 303 generates a data pattern (such as that shown in FIG. 6) according to the data value held by the data holding unit.

  The synthesizer 304 synthesizes the non-zero frequency components generated by the correction pattern generator 301 and the data pattern generator 303 to generate a pattern as shown in FIG. Note that the term “synthesis” here means addition.

  The inverse discrete Fourier transform unit 305 performs an inverse discrete Fourier transform on the non-zero frequency components synthesized by the synthesis unit 304 to generate a pattern in the spatial domain. The size of the pattern output by the inverse discrete Fourier transform unit 305 is determined by the number of samples handled in the inverse discrete Fourier transform, and is, for example, a size of 256 × 256 pixels, 512 × 512 pixels, 1024 × 1024 pixels, or the like.

  The binarization unit 306 binarizes the pattern generated by the inverse discrete Fourier transform unit 305 into a predetermined value a or −a using a predetermined value as a threshold value.

  The addition unit (or subtraction unit) 307 adds (or subtracts) the binarized pattern to all or a part of blocks of the Y (yellow) component in the original image represented by CMYK. The Y component is selected because the noise of the Y component is not visually noticeable. Therefore, for example, when the pattern is not easily noticeable, a component other than the Y component may be selected. Note that the binarization processing by the binarization unit 306 is performed because an experimental result has been obtained that the accuracy of digital watermark detection in the digital watermark detection apparatus is improved. Therefore, the binarization process is not necessarily required. In this case, the adding unit (or subtracting unit) 307 applies a pattern obtained by multiplying the pattern output from the inverse discrete Fourier transform unit 305 by a predetermined coefficient to the original image. Of these, the Y component is added (or subtracted) to all or a part of the blocks. Further, it is presumed to be due to the characteristics of the optical system or electrical processing system of the camera, but the pattern output by the inverse discrete Fourier transform unit 305 or a binarized pattern is also slightly added to the other components of the Y component ( (Or subtraction) in comparison with the case where the pattern output from the inverse discrete Fourier transform unit 305 or the binarized pattern is added (or subtracted) to only the Y component. Since an experimental result has been shown that the accuracy is improved, a pattern output by the inverse discrete Fourier transform unit 305 or a binarized pattern is also slightly added to some or all blocks of other components of the Y component (or Subtraction).

  Next, the operation of the digital watermark insertion apparatus according to the present embodiment will be described with reference to FIG.

  First, the correction pattern generation unit 301 generates a correction pattern (as shown in FIG. 2) (step S401). Next, the data pattern generation unit 303 generates a data pattern (such as shown in FIG. 6) determined by the data value held by the data holding unit 302 (step S402). Next, the pattern synthesis unit 304 synthesizes the patterns generated by the correction pattern generation unit 301 and the data pattern generation unit 303 (step S403). Next, the inverse discrete Fourier transform unit 305 performs an inverse discrete Fourier transform on the pattern (as shown in FIG. 7) obtained by the synthesis unit 304 to generate a pattern in the spatial domain (step S404). Next, the binarization unit 306 binarizes the pattern in the spatial domain generated by the inverse discrete Fourier transform unit 305 (step S405). Finally, the addition unit (or subtraction unit) 307 adds (or subtracts) the binarized pattern to the Y component of the original image and other components as necessary (step S406).

  The image into which the digital watermark is inserted is printed on a book, a magazine, or the like, or displayed on a screen of a homepage or the like, and waits to be photographed by a mobile phone camera or the like.

  Next, a digital watermark detection apparatus according to an embodiment of the present invention will be described.

  Referring to FIG. 10, the digital watermark detection apparatus according to the present embodiment includes a blocking unit 311, a discrete Fourier transform unit 312, an amplitude calculation unit 313, an amplitude integration unit 314, a correction value detection unit 315, a correction unit 316, and a data detection unit. 317.

  The blocking unit 311 blocks the B component in RGB of the digital watermark insertion image input by the mobile phone camera or the like. In the original image represented by CMYK, the pattern inserted into all or a part of the blocks of the Y (yellow) component is inserted into the B component when photographed with the RGB primary color filter camera. Because. Therefore, when the pattern is inserted into components other than the Y component, components other than the B component (specifically, components having a complementary color relationship with the component into which the pattern is inserted) are blocked. In addition, since a camera or the like captures an image printed on a book, a magazine, or the like, or an image displayed on a screen, an electronic watermark inserted image input by the camera or the like is an electronic watermark output by an electronic watermark inserting device. Compared to the inserted image, the origin is generally shifted, enlarged or reduced, rotated, and contrast and brightness are different. As described above, when the digital watermark is also inserted into the other components of the Y component, the other components of the B component are also blocked. The block is a block composed of a number of pixels equal to the number of pixels converted by the inverse discrete Fourier transform unit 305 of the digital watermark insertion apparatus. For example, a block of 256 × 256 pixels, 512 × 512 pixels, 1024 × 1024 pixels, or the like It is. Note that the number of pixels constituting the block created by the blocking unit 311 is equal to the number of pixels converted by the inverse discrete Fourier transform unit 305 of the digital watermark insertion apparatus, but the image input by the blocking unit 311 is a camera or the like. Therefore, the block generated by the blocking unit 311 is generally an enlarged or reduced block of the inverse discrete Fourier transform by the inverse discrete Fourier transform unit 305 of the digital watermark insertion apparatus.

  The discrete Fourier transform unit 312 performs discrete Fourier transform on all the blocks output by the blocking unit 311 and outputs all frequency components of all the blocks output by the blocking unit 311.

  The amplitude calculation unit 313 obtains the amplitude (absolute value) of the frequency component, which is generally a complex number output from the discrete Fourier transform unit 312, and outputs the amplitudes of all the frequency components of all the blocks output from the blocking unit 311.

  The amplitude integrating unit 314 adds the amplitude of each frequency component between all the blocks based on the amplitude of the frequency component output from the amplitude calculating unit 313.

  The correction value detection unit 315 detects the enlargement / reduction ratio and rotation angle of the image input by the blocking unit 311 based on the amplitude of each frequency component in the frequency domain output from the amplitude integration unit 314.

  For detection, first, the amplitude of each frequency component represented by orthogonal coordinates in the frequency domain is converted into the amplitude of each frequency component represented by polar coordinates. Then, as shown in FIG. 3, non-zero frequency components 201 "-1 to 201" -11 arranged in a straight line can be obtained. A straight line 202 ″ is obtained based on these non-zero frequency components 201 ″ -1 to 201- ″ 11, an enlargement / reduction ratio is obtained based on the slope of the straight line 202 ″, and the θ intercept of the straight line 202 ″ is obtained. The rotation angle is detected based on the value.

  The correction unit 316 corrects the position of the amplitude of each frequency component in the frequency region output from the amplitude integration unit 314 with the enlargement / reduction rate and rotation angle detected by the correction value detection unit 315, whereby the enlargement / reduction rate is 1. And the amplitude of each frequency component in the frequency domain when the rotation angle is 0 is obtained.

  The data detection unit 317 detects data based on the amplitude of each frequency component in the frequency domain when the enlargement / reduction ratio from the correction unit 316 is 1 and the rotation angle is 0. Data is detected by detecting a data pattern. Since the enlargement / reduction ratio and the rotation angle are corrected, the total amplitude corresponding to the data pattern appears at the candidate position of the predetermined radius. Accordingly, the summed amplitude after the correction of the enlargement / reduction ratio and the rotation angle is passed through a filter that passes the frequency component only at the candidate position. Then, for each of the circles 211 and 212, the total amplitude of the filter output is compared with a common threshold value that changes from a high value to a low value, and the total amplitude whose value exceeds the threshold value is a predetermined number (the circle 211). The change of the threshold value is stopped when 10 is reached for 14 and 14 for the circle 212 (see FIG. 11). At this time, the value of data embedded in the form of a digital watermark is detected from the pattern formed by the total amplitude exceeding the threshold.

  Next, the operation of the digital watermark insertion apparatus according to the present embodiment will be described with reference to FIG.

  First, the blocking unit 311 blocks the B component of the digital watermark insertion image taken by the camera or the like and other color components as necessary (step S411). Next, the discrete Fourier transform unit 312 performs discrete Fourier transform on each block generated by the blocking unit 311 (step S412). Next, the amplitude calculation unit 413 obtains the amplitude of each complex frequency component output from the discrete Fourier transform unit 312 (step S413). Next, the amplitude integrating unit 314 integrates the amplitude output from the amplitude calculating unit 313 for each frequency between blocks, and obtains a total amplitude for each frequency (step 414). Next, the correction value detection unit 315 obtains the enlargement / reduction ratio and rotation angle of the digital watermark inserted image photographed by the camera or the like based on the total amplitude for each frequency output from the amplitude integrating unit 314 (step S415). Next, the correction unit 316 corrects the position of the total amplitude for each frequency in the frequency domain with the enlargement / reduction ratio and rotation angle obtained in step S415 (step S416). Finally, the data detection unit 317 detects data embedded in the form of a digital watermark based on the total amplitude for each frequency output from the correction unit 316 (step S417).

  The digital watermark insertion apparatus shown in FIG. 8 can be realized by hardware, but the computer reads and executes a program for causing the computer to function as the digital watermark insertion apparatus shown in FIG. Can also be realized. Similarly, the digital watermark detection apparatus shown in FIG. 10 can be realized by hardware, but the computer reads and executes a program for causing the computer to function as the digital watermark detection apparatus shown in FIG. Can also be realized.

It is a figure which shows the pattern in the frequency domain of the digital watermark by a prior art example. It is a figure which shows the pattern in a frequency domain for detecting the expansion / contraction rate and rotation angle of an image by embodiment of this invention. It is a figure which shows the pattern at the time of plotting the pattern shown in FIG. 2 on a polar coordinate. It is a figure for demonstrating the method for detecting the expansion / contraction rate of an image using the pattern shown in FIG. It is a figure for demonstrating the method for detecting the rotation angle of an image using the pattern shown in FIG. FIG. 4 is a diagram illustrating a pattern in a frequency domain for representing data according to an embodiment of the present invention. It is a figure which shows the pattern which match | combined the pattern shown in FIG. 2, and the pattern shown in FIG. 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. It is a figure for demonstrating the method of the data detection by embodiment of this invention. 3 is a flowchart illustrating a digital watermark detection method according to an embodiment of the present invention.

Explanation of symbols

301 correction pattern generation unit 302 data holding unit 303 data pattern generation unit 304 synthesis unit 305 inverse discrete Fourier transform unit 306 binarization unit 307 addition unit (subtraction unit)
311 Blocking unit 312 Discrete Fourier transform unit 313 Amplitude calculation unit 314 Amplitude integration unit 315 Correction value detection unit 316 Correction unit 317 Data detection unit

Claims (15)

A generation step of generating non-zero frequency components (hereinafter referred to as “ non-zero frequency components ”) in which a plurality of amplitudes arranged on a straight line in a frequency region viewed in polar coordinates are generated by non-zero frequency component generation means ;
An embedding step of embedding the non-zero frequency component in an original image by embedding means ;
An electronic watermark insertion method comprising: The digital watermark insertion method according to claim 1,
The embedding step includes
Transforming the non-zero frequency component from the frequency domain to the spatial domain by frequency domain / spatial domain transforming means ;
Embedding the non-zero frequency component transformed into the spatial domain into the original image by means of embedding in the spatial domain;
An electronic watermark insertion method comprising: The digital watermark insertion method according to claim 1,
A digital watermark insertion method, further comprising a data insertion step of embedding a digital watermark representing data in the original image by data insertion means . Converting the image from the spatial domain to the frequency domain by the spatial domain / frequency domain conversion means ;
Non-zero frequency component detection means detects a plurality of non-zero frequency components arranged on a straight line in the frequency domain viewed in polar coordinates (hereinafter referred to as “ non-zero frequency component ”) from the image converted into the frequency domain . Detecting step to
A detection step of detecting an enlargement / reduction ratio and a rotation angle of the image based on the straight line by an enlargement / reduction ratio / rotation angle detection means ;
An electronic watermark detection method comprising: The digital watermark detection method according to claim 4,
The detecting step includes
When the rotation angle is represented by θ and the radius is represented by r in the frequency domain viewed in polar coordinates,
A step of detecting an enlargement / reduction ratio of the image based on a value when the straight line is expressed as θ = ar + b by an enlargement / reduction ratio detection unit ;
Detecting a rotation angle of the image based on a value of b when the straight line is expressed as θ = ar + b by a rotation angle detection unit ;
An electronic watermark detection method comprising: The digital watermark detection method according to claim 4,
By the correction means, and a correction step of correcting by the scaling factor and rotation angle are detected the image,
A data detection step of detecting a digital watermark representing the data from the corrected image by data detection means ;
An electronic watermark detection method, further comprising: Generating means for generating a non-zero frequency component (hereinafter referred to as “ non-zero frequency component ”) arranged in a straight line in a frequency domain viewed in polar coordinates;
Embedding means for embedding the non-zero frequency component in an original image;
An electronic watermark insertion apparatus comprising: The digital watermark insertion apparatus according to claim 7,
The embedding means includes
Means for transforming the non-zero frequency component from the frequency domain to a spatial domain;
Means for embedding the non-zero frequency component transformed into the spatial domain into the original image;
An electronic watermark insertion apparatus comprising: The digital watermark insertion apparatus according to claim 7,
A digital watermark insertion apparatus, further comprising data insertion means for embedding a digital watermark representing data in the original image. Means for transforming the image from the spatial domain to the frequency domain;
Detecting means for detecting, from the image converted into the frequency domain, a plurality of non-zero amplitude components arranged on a straight line in the frequency domain viewed in polar coordinates (hereinafter referred to as “ non-zero frequency component ”) ;
Detecting means for detecting an enlargement / reduction ratio and rotation angle of the image based on the straight line;
An electronic watermark detection apparatus comprising: The digital watermark detection apparatus according to claim 10, wherein
The detection means includes
When the rotation angle is represented by θ and the radius is represented by r in the frequency domain viewed in polar coordinates,
Means for detecting an enlargement / reduction ratio of the image based on a value when the straight line is expressed as θ = ar + b;
Means for detecting a rotation angle of the image based on a value of b when the straight line is expressed as θ = ar + b;
An electronic watermark detection apparatus comprising: The digital watermark detection apparatus according to claim 10, wherein
Correction means for correcting the image with the detected enlargement / reduction ratio and rotation angle;
Data detection means for detecting a digital watermark representing data from the corrected image;
An electronic watermark detection apparatus, further comprising:   The program for making a computer perform the method of any one of Claims 1 thru | or 6.   A computer-readable recording medium on which the program according to claim 13 is recorded.   A medium on which an image generated by the digital watermark insertion method according to any one of claims 1 to 3 is recorded.

Images