JP2010045453A - Apparatus, method, and program for electronic watermark detection - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an electronic watermark detecting apparatus for detecting watermark information even if time deviation arises in an image where a digital watermark is embedded. <P>SOLUTION: The electronic watermark detecting apparatus 2 computes a histogram for input images. The apparatus compares the computed histogram with a histogram of frames of a point before and after a scene change in an original image. The apparatus synchronizes scene changes based on the histogram comparison results. A frame counter measures the interval between scene change points for an input image, and estimates a frame rate of the input image based on a frame number of frames of a point before and after a scene change in an original image and the value of the frame counter. Then, the apparatus detects a watermark information 105 currently embedded in the input image 110, using the estimated frame rate. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

The present invention relates to an apparatus, a method, and a program for detecting a digital watermark embedded in content.

  A digital moving image can be easily made a high-quality copy at the digital signal level, and there is a possibility that the copy is unlimited without any copy inhibition or copy control. Therefore, in order to prevent illegal copying (copying) of a digital moving image, or to control the number of generations of copying by a regular user, a method of adding information for copy control to a digital moving image has been considered.

As a technique for superimposing other additional information on a digital moving image, a digital watermark is known. Digital watermarks are used for digital data such as audio, music, video, still images, etc., content copyright owner and user identification information, copyright owner rights information, content usage conditions, and usage Information such as secret information and duplication control information that are sometimes necessary is embedded in a state that cannot be recognized by human perception. Later, by detecting watermark information from the content as necessary, copyright protection including use control and copy control can be performed, or secondary use can be promoted.

Various methods have been proposed as a digital watermarking method, and a method using a spread spectrum technique is known as one of them. A technique has been proposed in which a watermark signal is created and embedded using a specific frequency component signal extracted from an image signal of an original image in which watermark information is embedded (for example, Patent Document 1). This method is highly resistant to geometric deformation (such as image cutout, physical scale conversion such as scaling). On the other hand, in recent years, the problem of unauthorized copying such as shooting a theater screen with a video camera has become apparent. It is known that a video image taken in this way has a temporal shift due to a rotational speed deviation of the projector in addition to geometric deformation. Therefore, there is a demand for a technique for stably detecting watermark information even when the frame rate changes due to a time lag.
JP 2006-254147 A

In the above prior art, embedded information is detected by accumulating for a long time with a specific code inversion pattern at the time of embedding at the time of detection, and in order to accurately detect watermark information, synchronization of the code inversion pattern is necessary. It becomes. When the frame rate changes due to a time lag, there is a problem that the detection accuracy of watermark information is not stable.

The present invention has been made to solve the above-described problem, and is a case where a time lag has occurred by preventing accumulation of accumulated pattern errors between embedding and detection due to a time lag. However, an object of the present invention is to provide a digital watermark detection apparatus, method and program capable of detecting watermark information more stably.

In order to solve the above-described problems, the present invention provides a digital watermark detection apparatus for detecting watermark information embedded in an input image, a histogram calculation means for calculating a histogram of a frame of the input image, and before embedding the watermark information. A histogram related to frames before and after each of a plurality of scene change points of the original image that is an image of the image, and a histogram about frames before and after each of a plurality of scene change points of the original image that is an image before embedding the watermark information, The level of correlation with a histogram relating to frames before and after the display time of the input image is compared, and a synchronization point of the input image having a correlation with each of a plurality of scene change points of the original image higher than a predetermined reference is obtained. The frame of the original image included between the synchronization point calculation means and the scene change point An estimation means for estimating an amount of change between the number of frames and the number of frames of the input image included between the synchronization points, and a polarity indicating a time-series change pattern of polarity when the watermark information is embedded in the original image The second period of the input image corresponding to the first period of the change pattern is estimated based on the amount of change, and the signal of the frame included in the second period has the same polarity as the first period. There is provided a digital watermark detection apparatus comprising: detection means for detecting the watermark information from a signal obtained by cumulative addition. Also provided are a method and a program used in the digital watermark detection apparatus of the present invention.

According to the present invention, it is possible to detect watermark information stably even if a time lag occurs.

Embodiments of the present invention will be described in detail with reference to the drawings. Hereinafter, the same reference numerals are given to the same components, and a part of overlapping description is omitted.

First, a digital watermark embedding device according to an embodiment of the present invention will be described. The digital watermark embedding device is information such as identification information of the copyright owner or user of the content, copyright holder rights information, content usage conditions, secret information necessary for the use, and copy control information (hereinafter referred to as watermark information). Embedded in the content.

FIG. 1 is a diagram illustrating a configuration of a digital watermark embedding apparatus 1 according to the present embodiment.

The digital watermark embedding apparatus 1 receives a digitized image signal of a moving image or a still image as an image in which watermark information is to be embedded (hereinafter referred to as “original image 100”). An embedded image 106 that is an image in which watermark information 105 is embedded in the original image 100 is generated and output. The image signal of the original image 100 may include both a luminance signal and a color difference signal, or may be only a luminance signal.

The digital watermark embedding apparatus 1 includes an embedding enlargement / reduction unit 10, a frequency component extractor 11, a watermark signal generator 12, and a watermark signal superimposing unit 13.

The embedded enlargement / reduction unit 10 reduces (or enlarges) the input original image 100 and outputs it to the frequency component extractor 11. Here, the embedded enlargement / reduction unit 10 may output a signal obtained by multiplying the original image 100 by one, that is, without performing enlargement or reduction processing.

The frequency component extractor 11 extracts a specific frequency component from the reduced (or enlarged) signal of the original image 100 and outputs it to the watermark signal generator 12 (hereinafter, the output signal from the frequency component extractor 11 is referred to as “ Specified frequency component)). The frequency component extractor 11 has a frequency domain digital filter for extracting a specific frequency component, for example, a low pass filter or a high pass filter having a predetermined cutoff frequency, or a band pass filter having a predetermined pass band center frequency. . Here, the filter for extracting the specific frequency component to be used needs to extract a specific frequency component equivalent to that at the time of detection. Further, the frequency component extractor 11 may extract all the frequency components of the input signal as specific frequency components.

  The watermark signal generator 12 generates a watermark signal 101 based on the input specific frequency component and the watermark information 105 and outputs it to the watermark signal superimposing unit 13. A method for generating the watermark signal 101 will be described.

The watermark information 105 is a signal sequence of “1” or “0” of a digital signal, and has a plurality of bits of data. First, the embedding position in the frame of the original image 100 is set for each bit of the watermark information 105. Each bit of the watermark information 105 is embedded at a position in a different frame. The setting of the embedding position of the watermark information 105 is controlled by a phase shift amount by a single or a plurality of digital phase shifters. A watermark signal 101 obtained by multiplying the amplitude of the specific frequency component whose phase is shifted by a coefficient corresponding to the value of each bit (“1” or “0”) is generated and output to the watermark signal superimposing unit 13. The process of controlling the amplitude of the specific frequency component is performed by a single or a plurality of exclusive OR circuits or digital multipliers.

FIG. 2 is a diagram showing an example of the state of phase shift. 2 and 3 exemplarily show a signal of one line in the two-dimensional image data, a one-dimensional signal is drawn. In both figures, the horizontal axis indicates the position in the image, and the vertical axis indicates the value of the image signal (for example, the luminance value).

The phase of the original image 100 input to the watermark signal generator 12 is shifted while maintaining the waveform of the specific frequency component signal. The embedding position of the watermark information 105 in the two-dimensional image data is controlled by the phase shift amount. The watermark signal generator 12 uses a predetermined shift amount. The shift amount is set in advance according to the information amount of the watermark information 105 to be embedded, and is given to the watermark signal generator 12. The sign and size of the coefficient to be multiplied by the specific frequency component signal whose phase is shifted are determined according to the activity representing the complexity of the original image 100 or the like. For example, the larger the activity, the larger the coefficient is set. Note that the coefficient differs depending on whether the bit value is “1” or “0”.

The watermark signal superimposing unit 13 is a digital adder, and superimposes the input watermark signal 101 on the original image 100 to generate an embedded image 106. The watermark signal superimposing unit 13 embeds the watermark signal 101 in the original image 100 while inverting the sign (+ −) of the watermark signal 101 with a predetermined pattern. For example, superimposition is performed on each frame of the original image 100 with a polarity change pattern such as inverting the sign of the watermark signal 101 every 10 frames. This is to facilitate separation of the image signal component (the signal component of the original image 100) and the watermark signal 101 from the image in which the watermark information is embedded at the time of detection. A frame in which the watermark signal 101 is superimposed with the same code (polarity) is a frame included in the same period (first period).

The specific frequency component signal extracted by the frequency component extractor 11 may exist in a plurality of channels. In this case, the watermark signal 101 generated based on the specific frequency components of the plurality of channels is respectively converted into the watermark signal superimposing unit 13. Is superimposed on the original image 100.

The frequency component extractor 11 extracts a specific frequency component signal of the embedding target image signal (original image 100) shown in FIG. FIG. 3B is an extracted specific frequency component signal (image of a specific frequency component).

The specific frequency component signal generates two different phase shift signals that are phase-shifted by a predetermined shift amount determined in advance by the watermark signal generator 12 (two phase shifters). Next, the generated phase shift signal is multiplied by a coefficient representing the 0th bit or the 1st bit of the watermark information 105. For example, if the watermark information 105 is “0”, the phase shift signal is multiplied by −1, and if it is “1”, the phase shift signal is multiplied by +1. The reason why the watermark signal generator 12 is two phase shifters is an example in which the generated phase shift signal is 2 bits.

FIGS. 3C and 3D are diagrams showing examples of phase shift signals when the watermark information 105 to be embedded is 2-bit information (1, 1). There are four combinations of 2-bit information: (0, 0) (0, 1) (1, 0) (1, 1). As shown in FIG. 3 (c), as a signal corresponding to the first bit, a signal having a specific frequency component signal having a + sign corresponding to 1 and a phase shifted by α in the X direction is generated. Further, as shown in FIG. 3 (d), as a signal corresponding to the second bit, a specific frequency component signal is a + sign corresponding to 1, and a signal obtained by shifting the signal shifted in the X direction by 2α is obtained. Generate. Since the phase of the image corresponds to the position of the image, the phase shift represents the movement of the position within the screen. 3C and 3D, the position of the signal differs between the specific frequency component signal and the phase shift signal 1 due to the phase shift, and the position of the leftmost peak of the signal differs (the position of the peak position). The difference is due to the phase being shifted). Note that the positive / negative relationship between the signal and the sign and the phase shift amount are examples, and various modifications may be made. Further, although the phase shift amount is exemplified as a one-dimensional direction shift, it indicates a position shift on a two-dimensional screen.

After that, the embedded image signal (embedded image 106) generated by the watermark signal superimposing unit 13 adding the phase shift signal multiplied by the factor for bit representation to the original image 100 is shown in FIG. Shown in solid line. The broken line in FIG. 3E shows the embedding target image signal before being superimposed, and the phase shift signal shown in FIGS. 3C and 3D.

As described above, the image signal (hereinafter referred to as “embedded image”) 106 in which the watermark information 105 is embedded is output.

Hereinafter, first to third embodiments of a digital watermark detection system for detecting watermark information from an input image (an image in which watermark information is embedded) input via a recording medium or a transmission medium will be described with reference to the drawings. explain.

[First embodiment]
A digital watermark detection system according to the first embodiment will be described.

FIG. 4 is a diagram showing a configuration of the digital watermark detection apparatus 2 of the present embodiment. The digital watermark detection apparatus 2 includes a feature amount storage unit 30, a histogram calculator 31, a histogram comparator 32, a synchronization point calculator 33, a frame rate estimator 34, a frame counter 35, and a digital watermark detector 36.

The digital watermark detection device 2 separates the components of the watermark signal 101 from the input image 110 (with the digital watermark embedded), and detects the watermark information 105.

The feature amount storage unit 30 stores an original image histogram 103 that is a histogram before and after the scene change point of the original image and a frame number 104 of the corresponding scene change point as the feature amount of the original image. In this embodiment, an example using a histogram based on luminance will be described.

The histogram calculator 31 calculates an input image histogram 107 from the input image 110 that has been input. Note that it is also possible to preliminarily estimate points at which scene changes have occurred in frames before and after the display time of the input image 110, and obtain histograms of frames before and after the estimated points. The histogram comparator 32 compares the input image histogram 107 obtained by the histogram calculator 31 with the original image histogram 103 held in advance in the feature amount storage unit 30.

The synchronization point calculator 33 compares the original image histogram 103 with the input image histogram 107 and obtains a point of the input image 110 having a high correlation with the original image histogram 103 as a synchronization point among the points of the input image.

Based on the degree of coincidence, a synchronization point that is a point that is synchronized with the original image histogram 103 in the input image 110 is obtained. The original image histogram 103 of each frame before and after (or the average of a plurality of frames in the same scene) and the input image histogram 107 of the frame of the input image 110 are compared.

In this embodiment, the original image histogram 103 of each frame (or the average of a plurality of frames in the same scene) before and after the scene change point of the original image and the input image histogram 107 of each frame of the input image 110 are compared. . Next, based on the comparison result (degree of correlation) of both histograms, the scene change point of the input image 110 corresponding to the scene change point of the original image is obtained as a synchronization point. Detailed processing will be described later.

The frame counter 35 counts the number of frames of the input image 110 included between the synchronization points of the input image 110 obtained by the synchronization point calculator 33 (synchronization point interval).

The frame rate estimator 34 calculates the number of frames included in the scene change point interval of the original image 100 and the frame included in the synchronization point interval of the input image 110 from the frame number 104 and the number of frames counted by the frame counter 35. Estimate the amount of change with the number of. In the present embodiment, the frame rate change rate α is estimated as the change amount. The frame rate change rate α indicates a difference in the number of frames included in the synchronization point interval synchronized with the scene change point interval. This is because the number of frames included between the input image 110 and the original image 100 even if the scene change point is synchronized due to a time lag caused by individual differences in the rotation speed of the projector. This is because a difference occurs. Note that the difference in the number of frames may be used as the amount of change.

The digital watermark detector 36 separates the watermark signal 101 from the input image 110 using the frame rate change rate α estimated by the frame rate estimator 34 and detects the watermark information 105. Detailed configuration and operation will be described later.

FIG. 5 is a flowchart showing the processing of the digital watermark detection apparatus 2.

First, the histogram calculator 31 calculates the input image histogram 107 (step S31). Next, in the histogram comparison, the histogram comparator 32 compares the original image histogram 103, which is a histogram before and after the scene change point of the original image, with the input image histogram 107 before and after the display time of the input image 110 (step S32). ). From the comparison result, the synchronization point of the input image 110 synchronized with the scene change point of the original image 100 is obtained (step S33). The synchronization point indicates the frame number of the input image at a point where the histogram before and after that is highly correlated with the histogram before and after the scene change point of the original image. The frame number is determined by the synchronization point calculator 33 from the number of frames included in the synchronization point interval counted by the frame counter 35.

Next, the frame rate change rate α between the input image 110 and the original image 100 is estimated (step S34). First, the frame counter 35 measures the number of frames of the input image 110 included between the synchronization points of the input image 110 (synchronization point interval). The number of frames included between scene change points of the original image 100 (scene change point intervals) is obtained from the frame number 104 stored in the feature amount storage unit 30. The frame number is stored in advance in the feature amount storage unit 30. The number of frames can be obtained by subtracting the frame number 104 of the first frame after a certain scene change point from the frame number 104 of the first frame after the next scene change point. The frame rate change rate α between the input image 110 and the original image 100 is estimated from the number of frames included in the scene change point interval and the number of frames included in the synchronization point interval synchronized with the scene change point interval.

Next, using the estimated frame rate, the watermark information 105 embedded in the input image 110 is detected, the watermark information is output, and the process is terminated (step S35). A detailed method for detecting the watermark information 105 will be described later.

A method in which the histogram comparator 32 compares the original image histogram 103 and the input image histogram 107 (step S32) and a method in which the synchronization point calculator 33 obtains a synchronization point from the comparison result (step S33) are shown in FIGS. This will be described with reference to FIG.

6 to 8 are diagrams for explaining a method of comparing the original image histogram 103 and the input image histogram 107. In (a), the scene change point in the original image 100 is N (OBn, OAn). The original image histogram 103 of the frame OBn before the display time N at the sea change point N is shown. (B) shows the original image histogram 103 of the frame OAn after the display time N at the display time. A point of the input image 110 to be compared is a point M (IBm, IAm). (C) shows the input image histogram 107 of the frame IBm before the display time point M. (D) shows the input image histogram 107 of the frame IAm after the display time point M. In the present embodiment, an example in which a direction histogram is used as a histogram will be described. A direction histogram is a histogram using the gradient of pixel values between adjacent pixels. It is a histogram which uses the gradient direction of the brightness | luminance between the adjacent pixels of each pixel, and the magnitude | size of gradient as a frequency. 6 to 8 show an example in which four gradient directions (up, down, left, and right) are used as components.

FIG. 6 is a diagram illustrating an example in which the degree of coincidence is obtained between the histograms before and after the scene change point N and the point M. (E) shows a comparison result obtained by multiplying the original image histogram 103 of the frame OBn and the input image histogram 107 of the frame IBm for each direction component. (F) shows a comparison result obtained by multiplying the original image histogram 103 of the frame OAn and the input image histogram 107 of the frame IAm for each direction component. The multiplication for each direction component indicates, for example, that the frequency of the upward component of the input image histogram 107 is multiplied by the frequency of the upward component of the original image histogram 103. The other direction components are similarly calculated.

When the degree of matching of the histograms of the frames to be compared is high, the multiplication result is a large value. Conversely, if the degree of matching of the histograms is low, the multiplication result becomes a small value. A scene change point is a place where the difference in histogram is large before and after the scene change point. When the degree of matching between the histograms before and after points N and M is high as shown in FIG. 6, it can be estimated that point M and scene change point N are synchronized. In this way, the synchronization point can be obtained from the degree of coincidence between the original image histogram 103 and the input image histogram 107 based on the magnitude of the multiplication result.

FIG. 7 is a diagram showing an example in which the degree of matching between histograms is obtained by a method different from that in FIG. The degree of matching of the histograms before the scene change point N and after the point M is obtained. Similarly, it is a diagram illustrating an example in which comparison is performed for obtaining the degree of matching between histograms after the scene change point N and before the point M. FIG. (E) shows a comparison result obtained by multiplying the original image histogram 103 of the frame OAn and the input image histogram 107 of the frame IBm for each direction component. (F) shows a comparison result obtained by multiplying the original image histogram 103 of the frame OAn and the input image histogram 107 of the frame IBm for each direction component.

As described above, when the degree of matching of the histograms is high, the multiplication result becomes a large value. As shown in FIG. 7, when the degree of matching of the histograms before and after points N and M is high, the degree of matching of histograms before scene change point N and after point M The degree of matching between the histograms after the point N and before the point M is lowered. In this case, it can be estimated that the point M is synchronized with the scene change point N.

As described above, the synchronization point of the input image 110 synchronized with the scene change point can be obtained from the magnitude of the result obtained by multiplying the histogram for each direction component.

FIG. 8 is a diagram showing an example in which the degree of coincidence between histograms is obtained by a method different from the method shown in FIG. 6 and FIG. (E) shows a comparison result obtained by dividing the original image histogram 103 of the frame OBn and the input image histogram 107 of the frame IBm for each direction component. (F) shows a comparison result obtained by dividing the original image histogram 103 of the frame OAn and the input image histogram 107 of the frame IAm for each direction component.

The higher the degree of matching of the histograms of the frames to be compared, the closer the division result is to 1. Conversely, the lower the degree of matching of histograms, the farther the division result is from 1. As shown in FIG. 8, when the degree of matching between the histograms before and after points N and M is high, it can be seen that the division result of each component is a value close to 1. In this case, it can be estimated that the point M and the scene change point N are synchronized. In this way, the degree of coincidence between the original image histogram 103 and the input image histogram 107 can be calculated from the division result. Note that if the compared histograms are similar, the division result is considered to be a value close to 1. Therefore, the statistics (for example, average or variance) of the division result may be used. In this case, when the average value of the division results is close to 1 or when the variance is small, it can be estimated that the compared histograms are similar.

It is also conceivable to calculate the degree of coincidence of histograms by combining the calculation methods of FIGS. 6 to 8 described here. In this example, multiplication and division are used to calculate the degree of histogram matching, but other methods such as addition and subtraction can also be used.

As described above, the original image histogram 103 and the input image histogram 107 are compared, a plurality of pairs of frame numbers having a high correlation of the histogram are obtained, and a synchronization point is obtained.

A method (step S34) in which the frame rate estimator 34 estimates the frame rate of the input image 110 based on the calculated synchronization point will be described with reference to FIG.

FIG. 9 is a diagram for explaining a method for estimating the frame rate of the input image 110. FIG. 9A shows a frame configuration of the original image 100. FIG. 9B shows a frame configuration of the input image 110. Both arrows indicate the display time.

The scene change points in the original image 100 are N (OBn, OAn), N + 1 (OBn + 1, OAn + 1), and N + 2 (OBn + 2, OAn + 2). The synchronization points of the input image 110 are M (IBm, IAm), M + 1 (IBm + 1, IAm + 1), and M + 2 (IBm + 2, IAm + 2). The synchronization point (scene change point in the input image 110) is obtained by comparison of histograms shown in FIGS.

The scene change point N (OBn, OAn) indicates that it is a scene change point between the frame with the frame number OBn and the frame with the number OAn. The same notation is used for other points.

The scene change point interval O1 between the scene change points N and N + 1 in the original image 100 is an OBn + 1−OBn frame from the frame number. Similarly, the scene change point interval O2 between the scene change points N + 1 and N + 2 is OBn + 2−OBn + 1 frame from the frame number.

The synchronization point interval I1 between the synchronization points M and M + 1 is an IBm + 1-IBm frame from the frame number. Similarly, the synchronization point interval I2 between the synchronization points M + 1 and M + 2 is IBm + 2−IBm + 1 frame.

When the ratio of the scene change point intervals is calculated, in the original image 100, O2 ÷ O1 = (OBn + 2-OBn + 1) ÷ (OBn + 1−OBn), and in the input image 110, I2 ÷ I1 = (IBm + 2−IBm + 1) ÷ (IBm + 1−IBm). Become. Here, if the scene change point in the original image 100 corresponds to the scene change point in the input image 110, the ratio of the scene change point interval even if the above-described temporal attack is made. Are almost identical. This is because even if the frame rate changes between the original image 100 and the input image 110, the ratio of scene change point intervals does not change, and the ratio of scene change point intervals does not affect the frame rate. By using the ratio of the scene change point interval between the original image 100 and the input image 110 for scene change synchronization, it is possible to obtain a point that is more accurately synchronized with the scene change point. Note that it is also possible to synchronize the original image 100 and the input image 110 using only the result of the degree of matching of the histograms shown in the histogram comparison example.

In FIG. 9, it is assumed that the scene change point N and the synchronization point M are the scene change point N + 1 and the synchronization point M + 1, and the scene change point N + 2 and the synchronization point M + 2 are synchronized with each other.

For example, assume that the scene change point interval O1 (OBn + 1−OBn) in the original image 100 is 1000 frames, and the synchronization point interval I1 (IBm + 1−IBm) corresponding to O1 of the input image 110 is 1100 frames. Since the scene change point N and the synchronization point M are synchronized with each other, the frame rate change rate α of the input image 110 can be estimated to be α = 1100/1000 = 1.1.

FIG. 10 is a diagram showing a configuration of the digital watermark detector 36.

The digital watermark detector 36 includes an enlargement / reduction unit 27, a frequency component extractor 20, a first orthogonal converter 21, a combiner 22, a second orthogonal converter 23, a cumulative adder 24, and an estimator 25.

The enlargement / reduction unit 27 enlarges (or reduces) the input image 110 in which the digital watermark is embedded. Note that the enlargement / reduction unit 27 may extract the frequency component and the frequency component extraction unit 20 may extract all the frequency components. The enlargement / reduction unit 27 enlarges (or reduces) the input image 110 at the same enlargement / reduction ratio as the embedded enlargement / reduction unit 10 used in the digital watermark embedding apparatus 1.

The frequency component extractor 20 extracts a specific frequency component from the input image 110 enlarged (or reduced) by the enlargement / reduction unit 27. The frequency component extractor 20 is a digital filter in the same frequency region as the frequency component extractor 11, for example, a low pass filter or a high pass filter having a predetermined cut-off frequency, or a band pass filter having a predetermined pass band center frequency.

The first orthogonal transformer 21 performs orthogonal transformation processing such as orthogonal transformation on the signal having the specific frequency component extracted by the frequency component extractor 20. Similarly, an orthogonal conversion process is performed on the input image 110.

The synthesizer 22 generates a signal obtained by performing a complex synthesis of the two signals after orthogonal transformation. Here, the amplitude component after the first orthogonal transformation may be subjected to complex synthesis with a specific gravity on the phase by adjusting the amplitude.

The second orthogonal transformer 23 performs orthogonal transformation (inverse orthogonal transformation) on the synthesized signal that has been complex-synthesized by the synthesizer 22. The orthogonal transformation here must be paired with the transformation in the first orthogonal transformation. For example, when the first orthogonal transformer 21 performs a Fourier transform, the second orthogonal transformer 23 performs an inverse Fourier transform.

The cumulative adder 24 adjusts the polarity based on the frame rate change rate α estimated by the frame rate estimator 34 while synchronizing with the polarity inversion pattern when the watermark signal superimposing device 13 superimposes the watermark signal 101. Then, cumulative addition of signals after the second orthogonal transformation is performed. For example, when embedding is performed by inverting the sign of the watermark signal 101 for each frame, it is necessary to perform accumulation while inverting the code for each frame. Also, when the frame rate is changed (for example, α = 2 [times]), the polarity is reversed every two frames at the time of detection, and embedding and detection are performed by cumulatively adding one frame of the same polarity. Must be synchronized. Details will be described later.

The estimator 25 estimates the watermark information 105 from the signal after cumulative addition (the combined signal after conversion). As the watermark information 105, “1” or “0” of a digital signal is embedded.

A method for detecting an electronic watermark (electronic watermark detection step S35) will be described.

FIG. 11 is a flowchart showing how the digital watermark detector 36 detects the embedded watermark information 105 from the input image 110.

First, the enlargement / reduction unit 27 enlarges (or reduces) the input image 110 that has been input (step S40). At this time, the enlargement / reduction ratio of the embedded enlargement / reduction unit 10 is enlarged (or reduced) at the same magnification. In some cases, the enlargement / reduction ratio is one time.

Next, the frequency component extractor 20 extracts a signal component of a specific frequency component of the enlarged / reduced input image 110 (step S41). At this time, the same frequency region as the frequency component extracted by the frequency component extractor 11 is extracted. Note that all frequency components may be extracted as specific frequency components.

Next, the first orthogonal transformer 21 performs orthogonal transformation processing such as orthogonal transformation on the signal extracted in step S41 (step S42).

As in step S42, the first orthogonal transformer 21 performs orthogonal transformation processing such as orthogonal transformation on the input image 110 (step S47).

Next, the synthesizer 22 complex synthesizes the two signals after orthogonal transformation (step S43). Here, the amplitude component after the first orthogonal transform may be subjected to complex synthesis that adjusts the amplitude and places a specific gravity on the phase.

The second orthogonal transformer 23 performs orthogonal transformation (inverse orthogonal transformation) on the composite signal after complex synthesis (step S44). The orthogonal transformation performed here needs to be paired with the transformation performed by the first orthogonal transformer.

Next, the cumulative adder 24 performs cumulative addition of the signal after the second orthogonal conversion (step S45). The cumulative addition is performed based on the frame rate change rate α estimated by the frame rate estimator 34, while inverting the polarity in a synchronized time series pattern when the watermark signal superimposing unit 13 embeds the watermark signal 101. . A detailed description of synchronization of time-series patterns will be described later.

Next, the watermark information 105 is estimated and output (step S45).

A method for estimating the watermark information 105 in step S45 will be described with reference to FIGS.

As shown in FIG. 2, while shifting the phase of the synthesized signal after the complex synthesis by the synthesizer 22, the correlation with the converted synthesized signal that has not been phase-shifted is obtained. Note that a phase-only correlation method or the like is used as a method for calculating the correlation (similarity) between two image signals.

FIG. 12 is a diagram illustrating the relationship between the cross-correlation value and the phase shift amount. As shown in FIG. 10, the peak of the cross-correlation value appears at a certain phase shift amount position. The polarity (positive / negative) of this peak represents the watermark information 105. The estimator 25 controls the phase shift amount continuously or stepwise, searches for the peak of the cross-correlation value output along with it, and estimates and detects the watermark information 105 from the polarity of the searched peak. The peak of the cross-correlation value takes either positive or negative value according to the watermark information 105. In FIG. 10, the watermark information 105 is determined to be “1” when positive, and the watermark information 105 is determined to be “0” when negative.

As described above, the specific frequency component signal is extracted from the input image 110, and the watermark information 105 is detected from the correlation result of the phase-only correlation between the specific frequency component signal and the image signal of the embedded image 106.

A method for synchronizing the patterns when the cumulative adder 24 performs cumulative addition (step S44) will be described.

FIG. 13 is a diagram for explaining a method in which the cumulative adder 24 cumulatively adds the signals after the second orthogonal transformation. FIG. 13A is a diagram illustrating a frame configuration of the original image 100. FIG. 13B is a diagram illustrating a frame configuration of the input image 110. In both figures, the arrow indicates the display time. A scene change point and a synchronization point connected by a broken line are points synchronized with each other.

Each period of the original image 100 is a polarity inversion pattern when the watermark signal 101 is superimposed, and the watermark signal 101 is superimposed with the same polarity in frames within the same period. The number of frames included in the first period is A frame. Similarly, the second period is a B frame, and the third period is a C frame.

The frame rate change rate estimated by the frame rate estimator 34 is α. The number of frames included in the first period of the input image 110 (period synchronized with the first period of the original image 100) is αA frame, the second period is αB frame, and the third period is αC frame. Therefore, in each period, it is necessary to superimpose signals with the same number of patterns and the same polarity as the number of frames included in the period of the original image synchronized with the period.

When α> 1, the same number of signals as the frames included in the synchronized period of the original image 100 are superimposed, for example, by not superimposing signals of predetermined frames included in the period. Further, when α <1, the same number of signals as the frames included in the synchronized period of the original image 100 are superimposed by, for example, accumulating signals of a predetermined frame included in the period a plurality of times.

In addition to the same number of frames to be superimposed, the inversion pattern in the input image 110 is estimated from the change rate α, and the frames included in the first period are obtained when the watermark signal 101 is superimposed on the original image. A method may be used in which αA frames are overlapped with the same inversion pattern as the inversion pattern and the patterns are synchronized.

In any method, it is possible to prevent accumulation of embedding and detection accumulated pattern errors due to a time lag between the original image 100 and the input image 110 from accumulating.

As described above, the polarity inversion pattern when the watermark signal superimposing unit 13 superimposes the watermark signal 101 is synchronized with the polarity inversion pattern when the cumulative adder 24 cumulatively adds the signals after the second orthogonal transformation. .

Thereby, even when the frame rate is changed due to a time lag caused by individual differences in the rotation speed of the projector, watermark information can be detected stably.

An operation example in which the digital watermark detection apparatus 2 detects the 2-bit watermark information 105 will be described with reference to FIGS. For simplicity, it is assumed that the enlargement / reduction ratio performed by the enlargement / reduction unit 27 is 1.

FIG. 14 is a diagram for explaining the operation of the digital watermark detection apparatus according to the first embodiment. Since one line of the signal of the two-dimensional image data is exemplarily shown, the one-dimensional signal is drawn.

When the watermark information 105 is detected from the input image 110, the frequency component extractor 20 extracts the specific frequency component signal shown in FIG. 14B from the input image signal shown in FIG. 14C and 14D are phase shift signals obtained by shifting the phase of the specific frequency component signal by a predetermined shift amount. A correlation value between the input image signal and the phase shift signal is obtained, and the watermark information 105 is determined from the peak of the correlation value. For example, if the correlation value peak is positive, the watermark information 105 is determined to be +1 (“1”), and if the correlation value peak is negative, the watermark information 105 is determined to be −1 (“0”). The When scaling (enlargement / reduction) is not performed on the input image 110, the input image signal has a shift amount (for example, the shift amount shown in FIGS. 3C and 3D) when watermark information is embedded. ) Has a correlation value peak at the same shift amount position. The embedded information estimator 25 detects the peak of the correlation value while controlling the phase shift amount. The watermark information 105 is estimated from the peak position.

FIG. 15 shows the correlation value peak search and watermark information detection operations when the watermark information 105 is (1, 1). As shown in FIG. 15, when there are two positive peaks of the correlation value other than the origin (a point where the phase shift amount is zero), (1, 1) as the watermark information 105 is detected.

FIG. 16 shows the correlation value peak search and watermark information detection operations when the watermark information 105 is (1, −1). As shown in FIG. 16, since the positive peak of the correlation value exists near the origin and the negative peak exists far from the positive peak from the origin, the watermark information 105 is (1, −1). Detected.

In addition, when the embedded image 106 is created by adding the specific frequency component signal to the original image 100, one of line, multiple lines, fields, fields, frames and frames, or You may use the system which inverts the polarity of a phase shift signal by these appropriate combinations, and the system which changes a phase shift amount.

In this embodiment, before the digital watermark detection apparatus 2 performs detection (when creating content, etc.), the feature amount (original image histogram 103, frame number 104) of the original image 100 is obtained in advance, and the feature amount storage unit 30 The example held in the above has been described. When the corresponding feature quantity is not stored in the feature quantity storage unit 30, the original image 100 is input to the digital watermark detection apparatus 2 at the time of detection, and the digital watermark detection apparatus 2 obtains the feature quantity from the original image 100. May be. In this case, the original image histogram 103 is calculated from the original image 100 as the feature amount, and the feature amounts of the frames before and after the scene change point are calculated. Also, it is necessary to calculate the frame numbers of the frames before and after the scene change point of the original image 100.

In addition, although the example of the histogram for comparison is one frame before and after the scene change, a histogram of a plurality of frames before and after the scene change may be obtained and the average may be compared.

Further, as a histogram of an image used by the histogram calculator 31 and the histogram comparator 32, a luminance histogram, a direction histogram, an edge histogram, or the like can be used. When any one of these histograms is used, a plurality of histograms may be used.

The luminance histogram is a histogram showing the distribution of how many pixels having which luminance value exist in the image.

An edge histogram is a histogram that uses the direction of an edge. The edge direction and intensity at each pixel in the image are calculated. It is a histogram obtained by dividing the direction of the edge into a plurality of classes and making the number of pixels assigned to the same class become the frequency of the class.

Depending on the assumed application (applied system), an optimal histogram can be used as appropriate. For example, when a large luminance change is assumed at a point other than the scene change point, the direction histogram and the edge histogram are considered to be effective. In addition, when a large luminance change other than the scene change point is not expected, it is considered effective to use the luminance histogram.

[Second Embodiment]
A digital watermark detection system according to the second embodiment will be described. The digital watermark detection apparatus of this embodiment performs a histogram calculation process on an image obtained by reducing the input image 110 in which the digital watermark is embedded.

FIG. 17 is a diagram showing a configuration of the digital watermark detection apparatus 3 of the present embodiment.

The digital watermark detection apparatus 3 of the present embodiment further includes a reduced image generator 37 as compared with the digital watermark detection apparatus 2 of FIG.

The reduced image generator 37 reduces the input image 110.

The histogram calculator 31 calculates a histogram from the reduced input image 110.

Various methods can be considered as a method of creating a reduced image by the reduced image generator 37, such as using a digital filter or thinning out an image in the horizontal direction or the vertical direction.

According to the digital watermark detection apparatus of the present embodiment, it is possible to reduce the amount of processing when calculating a histogram. In addition, the calculation of the histogram may be performed not on the entire screen of the image but on the center of the screen. In this way, not only the amount of processing is reduced but also the frame rate estimation performance is expected to be improved. This is because the input image 110 is likely to contain many original images 100 in the center of the screen. For example, when a captured image of a television screen is input to the detection system, components other than the original image 100 such as a television frame may be captured on the captured screen. There is little possibility that components other than the image 100 are captured, and the frame rate can be estimated more accurately.

[Third embodiment]
A digital watermark detection system according to a third embodiment will be described. The digital watermark detection system of this embodiment is an example in which histogram calculation and histogram comparison processing are performed on an image obtained by dividing the input image 110.

FIG. 18 is a diagram showing a configuration of the digital watermark detection apparatus 4 of the present embodiment.

The digital watermark detection apparatus 4 further includes a divided image generator 38 as compared with the digital watermark detection apparatus 2 of FIG.

The feature amount storage unit 30 of the digital watermark detection apparatus 4 holds a histogram (hereinafter, referred to as an original image division histogram) 108 of divided images of images obtained by dividing the original image 100 into a predetermined number in advance.

The divided image generation unit 38 divides the input image 110 into the same size and number as the divided areas into which the original image 100 is divided when generating the original image division histogram 108.

The histogram calculator 31 calculates a histogram of the divided image of the input image 110.

The histogram comparator 32 compares the histogram of the divided image of the input image 110 with the original image divided histogram 108.

The synchronization point calculator 33 obtains a point synchronized with the scene change point based on the comparison result of the histograms for each divided image. Although an example of two divisions is shown here, it is possible to increase the number of divisions. Further, there are various types of divisions such as dividing the input image in the horizontal direction, dividing in the vertical direction, and dividing in the horizontal and vertical direction.

It can be expected that the histogram and the scene change point interval ratio can be calculated more accurately by using the images divided in this way. Note that the optimal number of divisions may vary depending on the application assumed by the digital watermark detection system. When a large geometric deformation attack is assumed, an increase in the number of divisions is considered to lead to a decrease in performance. However, when a large geometric deformation is not assumed, it is considered that the performance is improved by an increase in the number of divisions.

The image processing apparatus according to each of the embodiments described above can be realized by using, for example, a general-purpose computer apparatus as basic hardware. In addition, even if a program-installable or executable format file is recorded and provided on a computer-readable recording medium such as a CD-ROM, CD-R, or DVD, it is provided by being incorporated in advance in a ROM or the like. May be. The program to be executed has a module configuration including each function described above.

1 is a diagram illustrating a configuration of a digital watermark embedding device according to a first embodiment. The figure which shows the example of the phase shift of a specific frequency component signal. The figure explaining operation | movement of a digital watermark embedding apparatus. 1 is a diagram illustrating a configuration of a digital watermark detection apparatus according to a first embodiment. The flowchart figure which shows the process of the digital watermark detection apparatus of 1st Embodiment. The figure explaining the method of comparing a histogram. The figure explaining the method of comparing a histogram. The figure explaining the method of comparing a histogram. The figure for demonstrating the method to estimate the frame rate of the input image. The figure which shows the structure of a digital watermark detector. 6 is a flowchart illustrating processing of the digital watermarking device according to the first embodiment. Diagram showing the relationship between cross-correlation value and phase shift amount The figure for demonstrating the method in which the cumulative adder 24 carries out the cumulative addition of the signal after 2nd orthogonal transformation. FIG. 6 is a diagram for explaining the operation of the digital watermark detection apparatus according to the first embodiment. The figure which shows the operation | movement of the peak search of a correlation value in case watermark information is (1, 1) and watermark information detection. The figure which shows the operation | movement of the peak search of a correlation value in case watermark information is (1, -1) and watermark information detection. The figure which shows the structure of the digital watermark detection apparatus of 2nd Embodiment. The figure which shows the structure of the digital watermark detection apparatus of 3rd Embodiment.

Explanation of symbols

100 ... Original image 110 ... Input image 101 ... Watermark signal 103 ... Original image histogram 107 ... Input image histogram 108 ... Original image division histogram 104 ... Frame number 105 ... Watermark information 106 ... Embedded image

DESCRIPTION OF SYMBOLS 1 ... Digital watermark embedding apparatus 10 ... Embedding expansion / reduction device 11 ... Frequency component extractor 12 ... Watermark signal generator 13 ... Watermark superimposing device

2, 3, 4 ... Digital watermark detection device

30 ... Feature value storage unit 31 ... Histogram calculator 32 ... Histogram comparator 33 ... Synchronization point calculator 34 ... Frame rate estimator 35 ... Frame counter 36 ... Digital watermark Detector 37 ... Reduced image generator 38 ... Split image generator

20 ... frequency component extractor 21 ... first orthogonal transformer 22 ... synthesizer 23 ... second orthogonal transformer 24 ... cumulative adder 25 ... estimator 27 ... enlarged Reducer

Claims (10)

In a digital watermark detection apparatus for detecting watermark information embedded in an input image,
Histogram calculation means for calculating a histogram of a frame of the input image;
Comparing the histogram of the frames before and after each of a plurality of scene change points of the original image, which is the image before embedding the watermark information, with the histogram of the frames before and after the display time of the input image, Synchronization point calculating means for obtaining a synchronization point of the input image having a correlation with each of a plurality of scene change points of the original image higher than a predetermined reference;
Estimating means for estimating an amount of change between the number of frames of the original image included between the scene change points and the number of frames of the input image included between the synchronization points;
Estimating a second period of the input image corresponding to a first period of a polarity change pattern indicating a time-series change pattern of polarity when embedding the watermark information in the original image based on the change amount; Detecting means for detecting the watermark information from a signal obtained by cumulatively adding the signals of the frames included in the second period with the same polarity as the first period;
An electronic watermark detection apparatus comprising: 2. The digital watermark detection according to claim 1, wherein the detection unit cumulatively adds signals of frames included in the second period and having the same number of frames as the number of frames included in the first period. apparatus. The estimation means obtains a change rate between the number of frames of the original image included between the scene change points and the number of frames of the input image included between the synchronization points as the change amount,
The said detection means estimates the said 2nd period from the pattern which multiplied the said change rate to the said polarity change pattern, and the said synchronization point which the said synchronization point calculation means calculated | required. Digital watermark detection device. 4. The electronic device according to claim 2, wherein the synchronization point calculation unit multiplies the histogram of the input image by the histogram of the original image, and determines that the correlation is higher as the multiplication result becomes larger. Watermark detection device. 4. The electronic device according to claim 2, wherein the synchronization point calculation unit divides the histogram of the input image by the histogram of the original image, and determines that the correlation is higher as the multiplication result is closer to 1. 5. Watermark detection device. The synchronization point calculation means obtains three or more synchronization points, the ratio of the number of frames of the input image included between the synchronization points, and the number of frames of the original image included between the scene change points. 4. The digital watermark detection apparatus according to claim 2, wherein a synchronization point that is synchronized with the scene change point is obtained using a ratio. The histogram calculation means divides the input image into a plurality of images,
The synchronization point calculation means compares the histogram of the original image and the histogram of the input image that are similarly divided,
The digital watermark detection apparatus according to claim 1, wherein the digital watermark detection apparatus is a digital watermark detection apparatus. The detection means includes
Extracting means for extracting a specific frequency component signal from the input image;
A first orthogonal transform unit for performing a first orthogonal transform on the specific frequency component signal and the input image to obtain a first orthogonal transform image;
A combining unit that combines the orthogonal transform images of the specific frequency component signal and the input image to generate a combined image;
A second orthogonal transformation means for performing a second orthogonal transformation paired with the first orthogonal transformation on the composite image to obtain a second orthogonal transformation image;
With
8. The digital watermark detection apparatus according to claim 1, wherein the watermark information is estimated from a signal obtained by cumulatively adding the second orthogonal transform image as a signal of the frame. In a digital watermark detection method for detecting watermark information embedded in an input image,
In a digital watermark detection apparatus for detecting watermark information embedded in an input image,
A histogram calculating step for calculating a histogram of a frame of the input image;
Comparing the histogram of the frames before and after each of a plurality of scene change points of the original image, which is the image before embedding the watermark information, with the histogram of the frames before and after the display time of the input image, A synchronization point calculating step for obtaining a synchronization point of the input image whose correlation with each of the plurality of scene change points of the original image is higher than a predetermined reference;
An estimation step for estimating a change amount between the number of frames of the original image included between the scene change points and the number of frames of the input image included between the synchronization points;
Estimating a second period of the input image corresponding to a first period of a polarity change pattern indicating a time-series change pattern of polarity when embedding the watermark information in the original image based on the change amount; A detection step of detecting the watermark information from a signal obtained by cumulatively adding signals of frames included in the second period with the same polarity as the first period;
An electronic watermark detection method comprising: In a digital watermark detection program for detecting watermark information embedded in an input image,
A histogram calculation function for calculating a histogram of a frame of the input image;
Comparing the histogram of the frames before and after each of a plurality of scene change points of the original image, which is the image before embedding the watermark information, with the histogram of the frames before and after the display time of the input image, A synchronization point calculation function for obtaining a synchronization point of the input image whose correlation with each of a plurality of scene change points of the original image is higher than a predetermined reference;
An estimation function for estimating the amount of change between the number of frames of the original image included between the scene change points and the number of frames of the input image included between the synchronization points;
Estimating a second period of the input image corresponding to a first period of a polarity change pattern indicating a time-series change pattern of polarity when embedding the watermark information in the original image based on the change amount; A detection function for detecting the watermark information from a signal obtained by cumulatively adding the signals of the frames included in the second period with the same polarity as the first period;
An electronic watermark detection program characterized by causing a computer to realize the above.

Images