JP2005236756A - Electronic watermark embedding device and electronic watermark detecting device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To embed electronic watermark information so that the electronic watermark information can be detected efficiently from an image, a portion of which is cut out to cause positional deviation and to detect the embedded electronic watermark information. <P>SOLUTION: An electronic watermark embedding device 100 generates an electronic watermark embedded transformation coefficient set A<SB>3</SB>from a discrete wavelet transformation coefficient set A<SB>2</SB>and a code string B<SB>2</SB>with a position marker embedded in electronic watermark information, and performs inverse discrete wavelet transformation of the electronic watermark embedded transformation coefficient set to generate electronic watermark embedded discrete image data A<SB>4</SB>having the position marker. An electronic watermark, detecting device generates a set of transformation coefficient, sets from the discrete image data by discrete wavelet, using oversampling, determines a code string for decoding electronic watermark, on the basis of reliability calculated from respective transformation coefficients and an electronic watermark code, detects the position marker from the code string, calculates the amount of a positional deviation, on the basis of the position marker and decodes the embedded information. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a digital watermark embedding device that embeds digital watermark information in image (including moving image) data and a digital watermark detection device that detects digital watermark information embedded in image data.

  The digital watermark embedding device using the discrete wavelet transform coefficient embeds the digital watermark information by modifying the value of the obtained transform coefficient according to a certain rule after performing discrete wavelet transform on the image data, and the digital watermark information is embedded. Digital watermark embedded image data is generated by inverse wavelet transform of the transform coefficient. Further, the digital watermark detection apparatus restores the embedded digital watermark information by performing discrete wavelet transform on the digital watermark embedded image data and then determining information embedded according to a certain rule from the obtained conversion coefficient (For example, refer to Patent Document 1).

Japanese Patent Laid-Open No. 2000-196856 (FIG. 1)

  However, when a part of a digital watermark embedded image is cut out and misaligned, the transform coefficient obtained by applying the discrete wavelet transform to the part of the image that is partly cut and misaligned is This is different from the transform coefficient obtained by applying the discrete wavelet transform to this image. This is because the discrete wavelet transform does not have shift invariance. For this reason, when a part of the digital watermark embedded image is cut out and misaligned, the embedded digital watermark information may not be detected.

  As shown in FIG. 1, the discrete wavelet transform is performed using recursive calculation stages (discrete wavelet decomposition stages) 11, 12,. 1, as shown in FIG. 2, the low-frequency component (discrete signal input in the first stage) obtained in the previous stage is subjected to high-frequency pass filter processing and low-frequency components. A pass filter process is performed, and further downsampling is performed to generate a high-frequency component and a low-frequency component with half the number of samples. 1 and 2 show the case of a one-dimensional signal, but in the case of a two-dimensional signal (image), the same processing is performed in order in the vertical direction and the horizontal direction. Four components (LL component, LH component, HL component, HH component) are generated from the low frequency component (discrete image signal input in the first stage) of the image data obtained in step (1). Here, L means a low-frequency component, H means a high-frequency component, and for example, the LL component is obtained by performing low-frequency pass filter processing in the vertical and horizontal directions and down-sampling in each direction. Means a set of transformation coefficients. In the next stage, a discrete wavelet transform is performed on the generated LL component. In the case of a two-dimensional signal (image), the number of samples of each component is a quarter of the number of samples of the LL component in the previous stage.

Further, in order to distinguish each component generated in each stage, for example, the LL component obtained in the Nth stage (also referred to as “stage N”) is referred to as “LLN component”. The position in the image corresponds to each conversion coefficient, and the position of the conversion coefficient of the component generated in stage N corresponds to the position when the resolution of the original image is 1/4 ^N. In addition, the size of one element of the LLN component is ^2N pixels in the vertical and horizontal directions in the original image. The case of the LHN component, the HLN component, and the HHN component is the same as the case of the LLN component.

  FIG. 3A shows the sampling position in each stage (stages 1 to 3) when the discrete wavelet transform is applied to the original image (stage 0), which is a one-dimensional signal, and FIG. Data obtained by shifting the signal one sample to the left is used as an original image (stage 0), and the sampling position in each stage (stages 1 to 3) when the discrete wavelet transform is applied to the original image is shown. As can be seen from the comparison between FIGS. 3A and 3B, the conversion coefficient obtained when the data shifted by one sample to the left is used as the original image is different from the conversion coefficient obtained when the data is not shifted. End up. With respect to the shifted original image, it is also possible to obtain a transform coefficient in which a digital watermark is embedded while shifting the position where the discrete wavelet transform is performed. However, this method cannot be said to be efficient in terms of computational complexity. Moreover, in order to obtain the same number of conversion coefficients, a process for complementing the left end or right end of the input sample is required. It was.

  In addition, reliability is usually given to the detected digital watermark information. When the digital watermark information is detected from a signal in which the digital watermark information is not embedded, the obtained information shows a statistically random value. Therefore, the reliability is set to a lower value as the obtained information is random, and to a higher value as the obtained information is biased. The presence / absence of a digital watermark is determined based on the reliability. However, in the conventional method, the obtained reliability must be absolutely evaluated, and there is a problem that the evaluation regarding the presence or absence of the digital watermark may not be properly performed depending on the conditions at the time of embedding and image processing after embedding. there were.

  Therefore, the present invention has been made to solve the above-described problems of the prior art, and the object of the present invention is to efficiently apply digital watermark information from an image that has been partially cut out and misaligned. It is an object to provide a digital watermark embedding device that can embed digital watermark information so that it can be detected, and a digital watermark detection device that can efficiently detect digital watermark information embedded by this digital watermark embedding device.

  Another object of the present invention is to provide a digital watermark detection apparatus that determines whether or not digital watermark information is embedded in image data based on a relatively efficient relative evaluation.

  The digital watermark embedding device of the present invention is obtained by encoding discrete wavelet transform unit that outputs a set of transform coefficients by performing discrete wavelet transform on input image data, and encoding information to be embedded as input digital watermark. An encoding and position marker insertion unit that generates a position marker-containing code string by inserting a position marker that is distinguishable from information to be embedded as a digital watermark into the code string, the conversion coefficient set, and the position marker included A digital watermark embedding unit that generates a set of transform coefficients with a digital watermark embedded with a position marker from the code sequence; and an inverse discrete wavelet transform of the digital watermark embedded transform coefficient set with a previous position marker, Inverse discrete to generate embedded discrete image data Those having a Eburetto conversion unit.

  Also, the digital watermark detection apparatus of the present invention is a discrete wavelet transform that generates and outputs a set of transform coefficient sets from the input discrete image data with embedded position marker-added watermark by discrete wavelet transform using oversampling. A digital watermark code detecting unit that determines a code string for decoding a digital watermark based on a reliability and a digital watermark code calculated from each of the transform coefficients included in the set of transform coefficient sets, and the digital watermark A position marker detection unit that detects a position marker from a code string output from the code detection unit, a code that calculates a positional deviation amount based on the detected position marker, and performs decoding using the obtained positional deviation amount And a digital watermark decoding unit that decodes information embedded as a digital watermark after determining the position of the column

  According to the digital watermark embedding device of the present invention, since the position marker is inserted into the code string of the information embedded as the digital watermark, the digital watermark detection is performed even from an image in which a part has been cut out and positional deviation has occurred. There is an effect that the digital watermark information can be embedded so that the digital watermark information can be efficiently detected by the apparatus.

  Further, according to the digital watermark detection apparatus of the present invention, a code sequence for decoding the digital watermark is determined based on the reliability calculated from each of the transform coefficients included in the set of discrete wavelet transform coefficient sets and the digital watermark code. Then, the amount of positional deviation is calculated based on the position marker detected from the code string, and the position of the code string to be decoded is determined using the obtained positional deviation amount, and then the information embedded as a digital watermark is decoded. As a result, the digital watermark information embedded by the digital watermark embedding apparatus can be efficiently detected.

<1 First Embodiment>
<1.1 Configuration>
FIG. 4 is a block diagram schematically showing the configuration of the still image digital watermark embedding apparatus 100 according to the first embodiment of the present invention. As shown in FIG. 4, the still image digital watermark embedding device 100 includes a discrete wavelet transform unit 110, an encoding and position marker insertion unit 120, a digital watermark embedding unit 130, and an inverse discrete wavelet transform unit 140. Have. The still image digital watermark embedding apparatus 100 receives the discrete image data A ₁ and the information B ₁ embedded as a digital watermark in the discrete image data A ₁ , and outputs the position marker-added digital watermark embedded discrete image data A ₄ . To do.

The discrete wavelet transform unit 110 receives the discrete image data A ₁ and outputs a discrete wavelet transform coefficient set (hereinafter referred to as “transform coefficient set”) A ₂ . The discrete wavelet transform unit 110 includes a multistage discrete wavelet decomposition (discrete wavelet transform) operation stage (an operation stage in which the discrete wavelet decomposition stage of FIGS. 1 and 2 is applied to a two-dimensional signal). A down-sampling unit that down-samples the filter and the signal that has passed through the high-frequency pass filter; and a down-sampling unit that down-samples the signal that has passed through the low-frequency pass filter and the low-frequency pass filter. The transform coefficient set is a sum set of LL components, HH components, LH components, and HL components obtained at each stage of the discrete wavelet transform. Here, L means a low-frequency component, H means a high-frequency component, and for example, the LL component is obtained by performing low-frequency pass filter processing in the vertical and horizontal directions and down-sampling in each direction. Means a set of transformation coefficients

Encoding and position marker insertion unit 120 inputs the information B ₁ to be embedded as an electronic watermark, and outputs the position marker containing binary codes (position marker containing electronic watermark code string) B _2. The digital watermark code sequence B ₂ with a position marker is obtained by binary-coding information to be embedded as a digital watermark, and a special code sequence of information (embedded as a digital watermark) called a position marker is added to the code sequence of information embedded as this digital watermark. A code string that can be distinguished from each other).

The digital watermark embedding unit 130 receives the transformation coefficient set A ₂ and the digital watermark code string B ₂ with a position marker as inputs, and outputs a transformation coefficient set A ₃ with a digital watermark embedded with a position marker. Inverse discrete wavelet transform unit 140 inputs the position marker containing watermarked transform coefficient sets A _3, and outputs the already discrete image data A ₄ embedded position marker containing watermark.

FIG. 5 is a block diagram schematically showing the configuration of the still image digital watermark detection apparatus 200 according to the first embodiment of the present invention. As illustrated in FIG. 5, the still image digital watermark detection apparatus 200 includes a discrete wavelet transform unit 210 using oversampling, a digital watermark code detection unit 220, a position marker detection unit 230, and a digital watermark decoding unit 240. And have. The still image digital watermark detection apparatus 200 receives the position marker-added digital watermark embedded discrete image data C ₁ and outputs information C ₅ embedded as a digital watermark.

The discrete wavelet transform unit 210 using oversampling is, for example, discrete image data C ₁ with embedded position marker-containing digital watermark output from the still image digital watermark embedding apparatus 100 (FIG. 4) (reference A _{4 in} FIG. ₄ ). , And a set C ₂ of transform coefficient sets obtained by oversampling is output. The discrete wavelet transform unit 210 using oversampling includes a multistage discrete wavelet decomposition operation stage. Each stage includes a high-frequency pass filter, an oversampling unit that oversamples a signal that has passed through the high-frequency pass filter, and a low-pass filter. A frequency pass filter and an oversampling unit that oversamples a signal that has passed through the low frequency pass filter are provided.

Electronic watermark code detecting section 220 receives as input a set C ₂ of transform coefficient sets, and outputs a set C ₃ position markers containing watermark code sequence. Position marker detection unit 230 receives as input a set C ₃ position markers containing watermark code sequence, detects and outputs the position marker information C ₄ from each code sequence. The digital watermark decoding unit 240 receives the position marker information C ₄ detected from each digital watermark code sequence and the set C ₃ of the digital watermark code sequence with the position marker, and outputs information C ₅ embedded as a digital watermark. .

Note that in the discrete wavelet transform unit 210 using oversampling, if calculation of transform coefficients using oversampling (that is, discrete wavelet transform) is performed from the first stage to the Nth stage, 2 ^N × 2 ^N position shifts. 4 ^N pieces of transform coefficient sets corresponding to the (positional deviation of 2 ^N as each of the vertical and horizontal directions) is obtained. In this case, the number of code sequences constituting the set of digital watermark code sequences output from the digital watermark code detection unit 220 is also 4 ^N , and the digital watermark decoding unit 240 uses the 4 ^N digital watermark codes. One code string that can read the digital watermark information from the string is selected, and the information embedded as the digital watermark is decoded.

<1.2 Operation>
First, the operation of the still image digital watermark embedding apparatus 100 will be described. In the still image digital watermark embedding device 100, the discrete wavelet transform unit 110 performs the processing procedure shown in FIG. 1 and FIG. 2 in the order in which the processing procedure shown in FIG. 1 and FIG. Are subjected to low frequency pass filter processing and high frequency pass filter processing, and the operation of downsampling is recursively repeated to generate the transform coefficient set A ₂ from the discrete image data A ₁ .

The encoding and position marker insertion unit 120 converts the information B ₁ to be embedded as a digital watermark into a binary code string, and inserts a position marker between the obtained digital watermark code strings. The position marker is expressed by a unique binary code string that does not appear in the digital watermark code string. When a part of the original image is cut out and the original image is displaced, this position marker is This is to allow the amount of positional deviation to be calculated by detecting the position marker. The encoding and position marker insertion unit 120, using the coding scheme as position markers do not appear in the code string generated, the information B ₁ to be embedded as electronic watermark information B ₁ to be embedded as an electronic watermark by encoding A code string is generated. By embedding a plurality of position markers in a code string of information to be embedded as a digital watermark, even if a part of the original image is cut out, some of the position markers can be detected in the remaining images.

The digital watermark embedding unit 130 embeds the digital watermark code sequence B ₂ with a position marker in the transform coefficient set A ₂ . For example, when the digital watermark code sequence B ₂ with a position marker is embedded in the LL component coefficient (the element of the LL component) in the transform coefficient set A ₂ , there is a method of embedding using the magnitude relationship with a reference value obtained from the peripheral LL component coefficient. is there. In this method, codes constituting the position markers containing electronic watermark code sequence B ₂ is larger than the reference value 1, so that 0 if less than the reference value, assigning a code to each transform coefficient.

Inverse discrete wavelet transform unit 140, a transform coefficient set A ₃ output from the digital watermark embedding unit 130, recursively position marker containing electronic watermark embedded discrete by integrating high and low frequency components in each stage generating image data _{a 4.}

  Next, the operation of the still image digital watermark detection apparatus 200 will be described. Unlike the normal discrete wavelet transform, the discrete wavelet transform unit 210 using oversampling in the still image digital watermark detection apparatus 200 performs oversampling. Oversampling is a technique in which the transform coefficients that have been discarded in the downsampling of the normal discrete wavelet transform are left without being discarded, and the discrete wavelet transform is similarly performed on these low frequency components.

6 is a diagram showing a configuration (in the case of a one-dimensional signal) of the discrete wavelet transform unit 210 using oversampling in FIG. 5, and FIG. 7 is a diagram of a discrete wavelet transform stage using oversampling in FIG. FIG. As shown in FIG. 6, in the discrete wavelet transform using oversampling, operation stages (that is, discrete wavelet transform stages using oversampling) 211, 212,... Are provided in multiple stages. In each stage of FIG. 6, as shown in FIG. 7, the low frequency component obtained in the previous stage (discrete signal input in the first stage) is subjected to high frequency pass filter processing and low frequency pass filter processing (down). Generates high and low frequency components without sampling). 6 and 7 show the case of a one-dimensional signal. In the case of a two-dimensional signal (image), the same conversion is performed in order in the vertical direction and the horizontal direction. As shown in FIG. 6 and FIG. 7, when the even-numbered conversion coefficient and the odd-numbered conversion coefficient are calculated for the two-dimensional signal (image) in the vertical direction and the horizontal direction, 1 By converting the degree, four types of LL component, LH component, HL component, and HH component are obtained. In each stage performs a discrete wavelet transform using the same manner oversampling respectively for four kinds of LL component obtained in the previous stage, respectively 4 ^N kinds of LLN component in the N stage LHN component, HLN Components and HHN components are obtained. Therefore, by performing the conversion from the first stage to the N stage, to obtain a 4 ^N types of transform coefficient sets corresponding to the displacement of 4 ^N Street (positional deviation of 2 ^N as each of the vertical and horizontal directions) be able to. However, regarding the transform coefficients obtained from the first stage to the (N−1) th stage, there is a common part between transform coefficient sets.

The downsampling in the discrete wavelet transform unit 110 of the digital watermark embedding device 100 for still images is 2 ^N samples in each of the vertical direction and the horizontal direction in terms of the resolution of the original image in the discrete wavelet transform stage of the Nth stage. Therefore, if a digital watermark is embedded in the conversion coefficient obtained by performing the conversion up to the Nth stage, the position of one sample from one sample to two ^N samples in each of the vertical and horizontal directions is set. The discrete wavelet transform is performed while shifting, and the digital watermark information can be correctly read from one of the obtained transform coefficient sets.

In the discrete wavelet transform unit 210 using oversampling of the digital watermark detection apparatus 200 for still images, if oversampling is performed from the first stage to the Nth stage, a transform coefficient set corresponding to a position shift of 2 ^N × 2 ^N is obtained. Since a set is obtained, a digital watermark can be read by selecting one of the obtained transform coefficient sets without performing discrete wavelet transform while shifting the position of the original image. However, the cut portion cannot be detected. Also, when the discrete wavelet transform is performed from the first stage to the Nth stage and the digital watermark is embedded in the obtained transform coefficient, the digital watermark can be read from the inside of the rectangular portion indicated by the dotted line 30 in FIG. Become. In FIG. 8, an outer frame rectangle 31 indicated by a solid line indicates an original image, an image obtained by cutting the outer side of the rectangle 32 indicated by a one-dot chain line and causing a position shift on the inner side of the rectangle 32 (an electronic watermark detection target) ).

  The digital watermark code detection unit 220 detects a digital watermark code string. The digital watermark code string detection method in the digital watermark code detection unit 220 is performed in the same manner as the digital watermark code string embedding method in the digital watermark embedding unit 130 of the digital watermark embedding apparatus 100 (FIG. 4). For example, when embedding of a digital watermark code sequence is performed using a magnitude relationship with a reference value that can be obtained from peripheral components, a reference that can similarly be used to detect a digital watermark code sequence from the peripheral components. This is done using a magnitude relationship with the value. Such detection processing is performed for each of all transform coefficient sets.

The digital watermark information is embedded in the transform coefficient generated in the Nth stage of the discrete wavelet transform unit 210 using oversampling, and the digital watermark embedded image is misaligned due to partial image clipping or the like. in this case, the electronic watermark information will be correctly decoded from only 4 ^N kinds of one code string in the electronic watermark code sequence. At this time, the transform coefficient in which the digital watermark is embedded appears only in one set of transform coefficient sets generated by oversampling.

If a digital watermark is forcibly decoded from a code string in which the digital watermark is not correctly decoded, a statistically random value is decoded. The reliability is determined based on this randomness, and the code string to be decoded is determined. For example, a method is adopted in which a plurality of position markers are inserted in a code string, position markers are detected from 4 ^N types of code strings, and a digital watermark is decoded from the code string in which the most position markers are detected. Can do. In this case, the reliability is represented by the number of detected position markers. In the case of a code string in which the digital watermark is not correctly decoded, the probability that the position marker is correctly detected is low due to the statistical randomness of the code.

  Note that the reliability determination method may be another method. For example, when a digital watermark is embedded in a magnitude relationship with the reference value, the absolute value of the difference from the reference value can be determined as the reliability. If the digital watermark is embedded relatively strongly (the absolute value of the difference is large), a code string in which the digital watermark is embedded relatively strongly can be determined as a code string to be decoded.

  The position marker detection unit 230 detects a position marker from each detection code string in order to obtain reliability and calculate a positional deviation amount. The digital watermark decoding device 240 determines the digital watermark code sequence to be decoded based on the reliability of each digital watermark code sequence represented by the number of detected position markers, etc., and calculates the positional deviation amount of the position marker. After the position of each code is determined using the amount of misalignment, decoding is performed.

<1.3 Effect>
As described above, according to the first embodiment, since the position marker is inserted into the digital watermark code string, even when the image is misaligned due to part of the image being cut off, By detecting the position marker, the amount of displacement can be calculated, and the conversion coefficient before the displacement can be obtained. For this reason, according to the digital watermark embedding apparatus 100 of the first embodiment, the digital watermark information can be embedded so that the digital watermark information can be efficiently detected from an image that is partially cut out and misaligned. Further, according to the digital watermark detection apparatus 200 of the first embodiment, digital watermark information embedded by the digital watermark embedding apparatus 100 can be detected efficiently.

  Further, according to the first embodiment, it is possible to determine a code string to be read based on the reliability of a digital watermark code string detected from each set of transform coefficients, and to decode the digital watermark information. For this reason, the amount of calculation in the digital watermark detection apparatus 200 can be significantly reduced compared to the amount of calculation in the case of employing a method of performing discrete wavelet transform while shifting the position of the original image.

ここで、ｘは、最初のステージで行われる演算量で、ステージを重ねる毎に４分の１ずつ減っていく。 For example, in the case of a conventional apparatus that performs conversion from the first stage to the Nth stage in the method of performing discrete wavelet transform while shifting the position of the original image, downsampling is performed every 2 ^N samples in terms of the resolution of the original image. Since this is performed, it becomes impossible to obtain a conversion coefficient when (2 ^N × 2 ^N −1) position shifts are performed. Therefore, the discrete wavelet transform must be performed while shifting the position by ^2N in the vertical direction and the horizontal direction, and the number of calculations given by the following equation (1) must be performed.

Here, x is the amount of calculation performed in the first stage, and decreases by a quarter each time the stages are overlapped.

On the other hand, when the discrete wavelet transform using oversampling is performed as in the digital watermark detection apparatus 200 of the first embodiment, the LL component, the HL component, the LH component, and the HH component are respectively obtained in each stage. Since the calculation is performed four times at a time, the number of times given by the following equation (2) may be calculated.

Since the order of the amount of calculation given by the equation (1) is 4 ^N and the order of the amount of calculation given by the equation (2) is N, the comparison between the equations (1) and (2) shows that In the case of this embodiment, it can be seen that the amount of calculation is greatly reduced.

<2 Second Embodiment>
<2.1 Configuration>
In the first embodiment, the case where the target of embedding / detecting the digital watermark information is a still image, but in the second embodiment, the digital watermark information is embedded / detected in the moving image. Explain the case.

FIG. 9 is a block diagram schematically showing the configuration of a moving image digital watermark embedding apparatus 300 according to the second embodiment of the present invention. Video electronic watermark embedding apparatus 300 includes a discrete video data D _1, as input information E ₁ to be embedded as an electronic watermark, and outputs the electronic watermark embedded moving image data D _4.

As illustrated in FIG. 9, the moving image digital watermark embedding apparatus 300 includes a frame storage unit (frame buffer) 310, a digital watermark embedding unit 320, and a moving image reconstruction unit 330. The digital watermark embedding unit 320 has the same configuration as the still image digital watermark embedding apparatus 100 shown in FIG. 4 of the first embodiment. Frame storage unit 310 inputs the discrete moving image data D _1, and save the video data in each frame, and outputs the frame D _2. The frame from the frame storage unit 310 is output at a timing based on the synchronization signal H ₁ from the digital watermark embedding unit 320. Moving picture reconstruction unit 330, an electronic watermark embedded frame D ₃ outputted from the electronic watermark embedding section 320 as input, and outputs moving image data D ₄ obtained by connecting the frame D _3.

FIG. 10 is a block diagram schematically showing a configuration of a moving image digital watermark detection apparatus 400 according to the second embodiment of the present invention. As illustrated in FIG. 10, the moving image digital watermark detection apparatus 400 includes a frame storage unit 410 and a digital watermark detection unit 420. The frame storage unit 410 performs the same operation as the frame storage unit 310 in the moving image digital watermark embedding apparatus 300 of FIG. The frame storage unit 410 receives the discrete moving image data F ₁ embedded with the position marker-containing digital watermark, stores the moving image data for each frame, and outputs the frame F ₂ . The frame F ₂ from the frame storage unit 410 is output at a timing based on the synchronization signal H ₂ from the digital watermark detection unit 420.

  FIG. 11 is a block diagram schematically illustrating the configuration of the digital watermark detection unit 420 according to the second embodiment. As shown in FIG. 11, the digital watermark detection unit 420 includes a discrete wavelet transform unit 421 using oversampling, a digital watermark code detection unit 422, a reliability and watermark code storage unit 423, and a decoded code string determination unit. 424, a position marker detection unit 425, and a digital watermark decoding unit 426. The decoding code string determination unit 424, the position marker detection unit 425, and the digital watermark decoding unit 426 constitute a digital watermark decoding device 427 (corresponding to a digital watermark decoding device 437 in FIG. 12 described later).

Discrete wavelet transform section 421 using the oversampling inputs the position marker containing electronic watermark embedded discrete image data F _1, and outputs a set F ₂ conversion coefficient set obtained by oversampling. The discrete wavelet transform unit 421 using oversampling performs the same operation as the discrete wavelet transform unit 210 (FIG. 5) using oversampling in the first embodiment.

Electronic watermark code detecting section 422 receives as input a set F ₃ for conversion coefficient set to calculate the respective electronic watermark code sequence and its reliability. The reliability and watermark code accumulating unit 423 accumulates the digital watermark code detected by the digital watermark code detecting unit 422 and the reliability F _4, and the accumulated code is obtained when all the frames used for detection are input. column to output the reliability F _5. Reliability and watermark code storage unit 423, After completion the processing of one frame, and outputs a synchronizing signal H ₂ to tell it. The decoded code string determination unit 424 receives each reliability and the code string F ₅ as input, determines a code string to be decoded, and outputs the code string F ₆ . Position marker detection unit 425 receives as input code sequence F ₆ to be decoded, (position of the position marker, reliability) position marker information and outputs a code string F _7. The digital watermark decoding unit 426 receives the position marker information and the code string as the input F ₇ and outputs the decoded digital watermark information F ₈ .

<2.2 Operation>
First, the operation of the moving image digital watermark embedding apparatus 300 will be described. In the electronic watermark embedding apparatus 300 for moving, cut into frame units in order of time series by the frame storage unit 310 discrete video data D ₁ that is input sequentially watermark frames D ₂ cut out from the moving image data The data is sent to the embedding unit 320. Frame D ₂ cut out is by the electronic watermark embedding section 320 which functions similarly to the first embodiment embedded watermark for still images described in the device 100 (FIG. 4), the electronic watermark information is embedded, the position It is sent to the moving image reconstruction unit 330 as a frame D ₃ with a digital watermark embedded with a marker. Moving picture reconstruction unit 330 as soon captures all frames D _3, when integrated frame in chronological order, to reconstruct the position marker containing electronic watermark embedded moving image data D _4. Note that the digital watermark information is basically the same for all frames. However, by embedding an index indicating the difference in the embedded contents at the same time, the moving image is divided into several sections to obtain different information. Can also be embedded.

Next, the operation of the moving image digital watermark detection apparatus 400 will be described. As illustrated in FIG. 10, the moving image digital watermark detection apparatus 400 cuts out frames in time-series order from the moving image data F ₁ with embedded position-marked digital watermark embedded by the frame storage unit 410 and cuts out the frame F _2. Are sequentially sent to the digital watermark detection unit 420 (reference numeral 430 in the third embodiment to be described later). Digital watermark detection unit 420 sequentially reads the digital watermark information from the frame, made a statistical judgment from the stored result, and outputs the electronic watermark information F _8.

Next, the operation of the digital watermark detection unit 420 will be described. The discrete wavelet transform unit 421 using oversampling operates in the same manner as the discrete wavelet transform unit 210 using oversampling of the still image digital watermark detection apparatus 200. As shown in FIG. 11, the discrete wavelet transform unit 421 using oversampling transforms the frame F ₂ cut out from the moving picture by discrete wavelet transform using oversampling, and transforms corresponding to the respective positional deviations. generating a set of coefficients set F _3.

The digital watermark code detection unit 422 simultaneously calculates the digital watermark code string and the detected code reliability F ₄ from each transform coefficient set F ₃ . There are various reliability levels depending on the embedding method. For example, if the digital watermark code is generated based on the magnitude relationship with the reference value calculated from the peripheral components and embedded, the absolute value of the difference from the reference value can be used as the reliability. The same reliability may be used in the still image digital watermark detection apparatus shown in the first embodiment. However, in the still image use, the difference from the reference value is considerably large when embedding the digital watermark. Without it, the judgment is difficult. In the case of moving images, data extracted from multiple frames can be accumulated and statistically determined, so even if the difference from the reference value is small within one frame, reliability is generated. can do.

Reliability and watermark code storage unit 423 stores an electronic watermark code is detected from each frame and its reliability F _4. The accumulation of reliability is performed so that the statistical characteristics of all frames used for detection can be expressed. For example, when the difference between the transform coefficient and the reference value is added together, the sum of the differences corresponding to the transform coefficients not embedded with the digital watermark is positive when the difference corresponding to the transform coefficient not embedded with the digital watermark is positive. Since it moves randomly between negative values, it approaches 0. Otherwise, the absolute value of the sum is a large value.

The decoding code string determination unit 424 determines a digital watermark code string for decoding the digital watermark information based on the reliability and the storage information F ₅ output from the watermark code storage unit 423. The accumulated value of the reliability of the code string obtained from the transform coefficient set from which the watermark information cannot be read is significantly larger than the accumulated reliability value derived from the transform coefficient set from which the watermark information can be read. Lower value. Therefore, a digital watermark code string to be decoded is determined based on the obtained reliability (for example, a transform coefficient set having the highest reliability).

Position marker detection unit 425 detects the position marker F ₇ from the bit stream F ₆ determining the decoding. The digital watermark decoding unit 426 calculates a positional deviation amount from the detected position marker, determines the position of each code using the calculated positional deviation amount, performs decoding, and is embedded as a digital watermark. and outputs the information _{F 8} was.

<2.3 Effects>
As described above, according to the second embodiment, since the position marker is inserted into the digital watermark code string, even if the image is misaligned due to part of the image being cut off, By detecting the position marker, the amount of displacement can be calculated, and the conversion coefficient before the displacement can be obtained. Therefore, according to the digital watermark embedding apparatus 300 of the second embodiment, the digital watermark information can be embedded so that the digital watermark information can be efficiently detected from an image that is partially cut out and misaligned. Also, according to the digital watermark detection apparatus 400 of the second embodiment, the digital watermark information embedded by the digital watermark embedding apparatus 300 can be detected efficiently.

  Further, in the digital watermark detection apparatus 400 of the second embodiment, the amount of calculation can be greatly reduced, as in the case of the digital watermark detection apparatus 200 of the first embodiment. That is, according to the second embodiment, the code sequence to be read can be determined based on the reliability of the digital watermark code sequence detected from each set of transform coefficients, and the digital watermark information can be decoded. The amount of calculation in the digital watermark detection apparatus 400 can be significantly reduced as compared with the amount of calculation in the case of employing a conventional method in which discrete wavelet transform is performed while shifting the position of the original image.

  Furthermore, in the case of moving images, the reliability obtained from each frame can be accumulated, so that each frame does not have to have a large degree of reliability, and as a result, an electronic image that can be read correctly while suppressing the effect on image quality. A watermark code string can be determined.

<3 Third Embodiment>
<3.1 Configuration>
In the second embodiment, the code string to be decoded is determined based on the processing results of all the frames of the moving image data. In the third embodiment, the processing results of some frames of the moving image data are used. An example of determining a code string to be decoded based on the above will be described. In the third embodiment, if it can be determined that there is a code whose reliability is prominent compared to other code sequences, the discrete wavelet transform by oversampling is stopped at that time, and this corresponds to a code sequence whose reliability is prominent. A case where only the transform coefficient set is calculated, the reliability of each code is accumulated, and the digital watermark information is restored will be described.

The moving image digital watermark embedding device of the third embodiment is the same as that of the second embodiment. In the moving image digital watermark detection apparatus of the third embodiment, the configuration of the digital watermark detection unit 430 is different from that of the second embodiment. As illustrated in FIG. 12, the digital watermark detection unit 430 according to the third embodiment receives a frame F ₂ cut out from a moving image and outputs information F ₈ embedded as digital watermark information.

  FIG. 12 is a block diagram schematically showing the configuration of the digital watermark detection unit 430 in the moving picture digital watermark detection apparatus according to the third embodiment of the present invention. As shown in FIG. 12, the digital watermark detection unit 430 includes a frame distribution unit 431, a discrete wavelet transform unit 432 using oversampling, a discrete wavelet transform unit 433 for calculating a specific transform coefficient set, a digital watermark A code detection unit 434, a reliability and watermark code storage unit 435, a digital watermark presence / absence determination unit 436, and a digital watermark decoding device 437 are provided.

The frame distribution unit 431 receives the control signal G ₁ from the digital watermark presence / absence determination unit 436 and the extracted frame F ₂ as input, and performs a discrete wavelet transform unit 432 using oversampling or a discrete transform for calculating a specific transform coefficient set. The input frame F ₁₁ is output to the wavelet transform unit 433.

The wavelet transform unit 432 using the oversampling has the same structure as the wavelet transform unit 421 using the oversampling of the second embodiment, and outputs the conversion coefficient set F _12. Discrete wavelet transform unit 433 for a specific transform coefficient set calculation frame F ₁₁ and the control signal G ₂ cut out, and outputs an specific transform coefficient set F _13.

The digital watermark code detection unit 434 receives the transform coefficient set F ₁₂ or F ₁₃ as an input, and outputs a digital watermark code string and a reliability F ₁₄ . Electronic watermark code detecting unit 434, the control signal G ₃ from the electronic watermark presence determining unit 436, may detect the code string of all transform coefficients set to determine whether to detect only a specific set.

The reliability and watermark code storage unit 435 receives the detected digital watermark code and the reliability F ₁₄ and outputs the stored code and reliability F ₁₅ . Reliability index and the watermark code storage unit 435, the control signal G ₄ from the electronic watermark presence determining unit 436, or stores the reliability for all transform coefficients set, accumulates only the reliability for a particular set To accumulate. Further, After completion processing of one frame to generate a synchronization signal H _2, it requests the next frame in the frame storage unit 410 (FIG. 10).

The digital watermark presence / absence determination unit 436 receives the accumulated reliability F ₁₅ as an input, determines whether or not to continue the reliability accumulation of all transform coefficient sets, and determines whether or not the frame distribution unit 431, the digital watermark code detection unit 434, Control signals G _{1 to} G ₅ are sent to the reliability and watermark code storage unit 435, the digital watermark decoding device 437, and the discrete wavelet transform unit 433 for calculating a specific transform coefficient set.

Electronic watermark decoding apparatus 437 starts the decoding of the watermark code After receiving the control signal G _5. As illustrated in FIG. 13, the digital watermark decoding apparatus 437 includes a decoded code string determination unit 438, a position marker detection unit 439, and a digital watermark decoding unit 440. The decoding code string determination unit 438, the position marker detection unit 439, and the digital watermark decoding unit 440 are configured in the same manner as the decoding code string determination unit 424, the position marker detection unit 425, and the digital watermark decoding unit 426 shown in FIG. Has been.

<3.2 Operation>
In FIG. 12, the frame sorting unit 431 uses a discrete wavelet transform unit 432 that uses oversampling of the frame F ₂ input to the digital watermark detection unit 430 according to a control signal or a discrete wavelet for calculating a specific transform coefficient set. Sort to the conversion unit 433.

The control signal from the digital watermark presence / absence determination unit 436 is set to perform oversampling at the start of moving image input, and then set to calculate only a specific transform coefficient set based on analysis of accumulated data. The discrete wavelet transform unit 432 using oversampling operates in the same manner as in the second embodiment. Discrete wavelet transform unit 433 for a specific transform coefficient set calculation performs control signals G ₂ receiving calculation of specified transform coefficient set from the electronic watermark presence determining unit 436. Each set of transform coefficients can be identified by whether odd-numbered transform coefficients are left or left even-numbered transform coefficients in the horizontal and horizontal directions at each discrete wavelet transform stage using oversampling ( FIG. 14). Therefore, instead of leaving all the conversion coefficients as in oversampling, the odd number in the vertical direction at each stage according to the specified sampling method (whether the even-numbered conversion coefficient or the odd-numbered conversion coefficient is left) The specific conversion coefficient set can be calculated by leaving either the odd number or the even number and leaving the odd number or the even number in the horizontal direction as well.

  Each stage performs the same processing on the LL component left in the previous stage. FIG. 15 shows a processing procedure for calculating a general wavelet transform and FIG. 16 for calculating a specific transform coefficient set. Based on the control signal, the digital watermark code detection unit 434 detects the digital watermark code for all transform coefficient sets, or detects the digital watermark code only for a specific transform coefficient set. The digital watermark code detection unit 434 calculates the reliability at the same time.

Reliability and watermark code storage unit 435 based on the control signal G _4, or accumulate reliability and sign of all transform coefficients set, or, to accumulate the reliability and code only for certain transform coefficient set. Based on the accumulated reliability, the digital watermark presence / absence determination unit 436 determines whether to continue oversampling further or to calculate only a set of transform coefficients corresponding to a code string from which digital watermark information can be decoded. . For example, when embedding digital watermark information based on the magnitude relationship with the reference value calculated from the peripheral components, the difference from the reference value is accumulated as an amount that simultaneously represents the code and reliability. Taking the sum of the absolute values of the accumulated values, it can be regarded as the reliability of each set of transform coefficients.

  By comparing these, it is possible to specify which digital watermark code string can be decoded. For example, it can also be determined that a code string that can be decoded has been identified at a stage where a difference between a code string having a large reliability and a reliability of another code string exceeds a certain threshold. If the determination is made, a control signal is generated and reliability is accumulated only for the specific conversion coefficient sequence. When the accumulation of reliability for all frames is completed, the result is passed to the digital watermark decoding apparatus 437. The digital watermark decoding device 437 determines a code string for decoding the digital watermark information from the accumulated reliability and code, detects a position marker, calculates a positional deviation amount, and all the code strings are the same as this position marker. Since the position shift has occurred, the position of each code is specified, and then the digital watermark information is restored.

<3.3 Effects>
As described above, according to the third embodiment, since the position marker is inserted into the digital watermark code string, even when the image is misaligned due to part of the image being cut off, By detecting the position marker, the amount of displacement can be calculated, and the conversion coefficient before the displacement can be obtained. Therefore, according to the digital watermark embedding device 300 of the third embodiment, it is possible to embed digital watermark information so that the digital watermark information can be efficiently detected from an image that has been partially cut out and misaligned. Also, according to the digital watermark detection apparatus 400 of the third embodiment, the digital watermark information embedded by the digital watermark embedding apparatus 300 can be detected efficiently.

  Further, in the digital watermark detection apparatus according to the third embodiment, the amount of calculation can be greatly reduced as in the case of the digital watermark detection apparatus 200 according to the first embodiment. That is, according to the third embodiment, the code sequence to be read can be determined based on the reliability of the digital watermark code sequence detected from each set of transform coefficients, and the digital watermark information can be decoded. The amount of calculation in the digital watermark detection apparatus can be greatly reduced compared to the amount of calculation in the case of employing a conventional method in which discrete wavelet transform is performed while shifting the position of the original image.

  Furthermore, in the case of moving images, the reliability obtained from each frame can be accumulated, so that each frame does not have to have a large degree of reliability, and as a result, an electronic image that can be read correctly while suppressing the effect on image quality. A watermark code string can be determined.

  Furthermore, according to the digital watermark detection apparatus of the third embodiment, it is not necessary to perform oversampling on all frames, so that the amount of calculation can be suppressed.

  Further, according to the digital watermark detection apparatus of the third embodiment, the digital watermark presence / absence determination unit 436 can be used to determine the presence / absence of digital watermark information. When oversampling is not performed and the presence / absence of digital watermark information is determined using only one transform coefficient set (transform coefficient set obtained by normal wavelet transform), absolute evaluation must be performed based on the obtained reliability. For example, when a certain threshold value is set for the reliability and it is attempted to determine the presence / absence of a digital watermark using this threshold value, there is a problem as to which value the threshold value is set to. In order to set this threshold value, it is necessary to statistically obtain data by conducting many experiments. On the other hand, if the reliability of the transform coefficient set for each position shift is obtained by oversampling, the reliability of the transform coefficient set that cannot be clearly detected and the reliability of the detectable set can be compared (relative evaluation). Thus, the presence / absence of digital watermark information can be more reliably determined.

It is a figure which shows the structure of the apparatus which performs a discrete wavelet transform of a one-dimensional signal. It is a figure which shows the structure of the discrete wavelet decomposition | disassembly stage of FIG. (A) is a figure which shows the sampling position in each stage at the time of carrying out discrete wavelet transform of the original image data which is a one-dimensional signal, (b) is the original data which shifted the one-dimensional signal 1 sample to the left. It is a figure which shows the sampling position in each stage at the time of setting it as image data and carrying out discrete wavelet transform. 1 is a block diagram schematically showing a configuration of a still image digital watermark embedding device according to a first embodiment of the present invention. FIG. 1 is a block diagram schematically showing the configuration of a still image digital watermark detection apparatus according to a first embodiment of the present invention. FIG. It is a figure which shows the structure (in the case of a one-dimensional signal) of the discrete wavelet transformation part using the oversampling of FIG. It is a figure which shows the structure of the discrete wavelet transformation stage using the oversampling of FIG. It is a diagram for explaining that a set of transform coefficients set corresponding to the displacement of the 2 ^N × 2 ^N by performing oversampling the first stage to the N stage is obtained. It is a block diagram which shows roughly the structure of the electronic watermark embedding apparatus for moving images based on the 2nd Embodiment of this invention. It is a block diagram which shows roughly the structure of the digital watermark detection apparatus for moving images concerning the 2nd Embodiment of this invention. It is a block diagram which shows roughly the structure of the digital watermark detection part in the digital watermark detection apparatus for moving images concerning the 2nd Embodiment of this invention. It is a block diagram which shows roughly the structure of the digital watermark detection part in the digital watermark detection apparatus for moving images concerning the 3rd Embodiment of this invention. It is a block diagram which shows roughly the structure of the digital watermark decoding apparatus in the digital watermark detection apparatus for moving images concerning the 3rd Embodiment of this invention. It is explanatory drawing of four sampling patterns which the discrete wavelet transformation part for the specific transformation coefficient set calculation in the digital watermark detection apparatus for moving images concerning the 3rd Embodiment of this invention performs. It is a flowchart which shows operation | movement of the discrete wavelet transformation part using the oversampling in the digital watermark detection apparatus for moving images concerning the 3rd Embodiment of this invention. It is a flowchart which shows operation | movement of the discrete wavelet transformation part for the specific transformation coefficient set calculation in the digital watermark detection apparatus for moving images concerning the 3rd Embodiment of this invention.

Explanation of symbols

100 still image digital watermark embedding device,
110 discrete wavelet transform unit,
120 encoding and position marker insertion unit,
130 digital watermark embedding unit,
140 Inverse discrete wavelet transform unit,
200 Still image digital watermark detection device,
210 discrete wavelet transform unit,
220 electronic watermark code detection unit,
230 position marker detector,
240 digital watermark decoding unit,
300 video watermark embedding device,
310 frame storage unit,
320 Digital watermark embedding unit,
330 moving image reconstruction unit,
400 electronic watermark detection apparatus for moving images,
410 frame storage unit,
420 digital watermark detection unit,
421 Discrete wavelet transform unit using oversampling,
422 a digital watermark detection unit;
423 reliability and watermark code storage unit;
424 decoded code string determination unit,
425 position marker detector,
426 digital watermark decoding unit,
430 electronic watermark detection unit,
431 Frame distribution part,
432 Discrete wavelet transform unit using oversampling,
433 Discrete wavelet transform unit for specific transform coefficient set calculation,
434 a digital watermark code detection unit;
435 reliability and watermark code storage unit,
436 electronic watermark presence / absence determination unit;
437 electronic watermark decoding apparatus;
438 decoding code string determination unit,
439 position marker detector,
440 A digital watermark decoding unit.

Claims (5)

A discrete wavelet transform unit that outputs a set of transform coefficients by performing discrete wavelet transform on input image data; and
A code that generates a code string with a position marker by inserting a position marker consisting of a code string that can be distinguished from information embedded as a digital watermark into a code string obtained by encoding information to be embedded as an input digital watermark And position marker insertion part,
A digital watermark embedding unit that generates a position marker-added digital watermark embedded transform coefficient set from the transform coefficient set and the position marker-added code string;
A digital watermark embedding comprising: an inverse discrete wavelet transform unit that generates discrete image data with a digital watermark embedded with a position marker by performing an inverse discrete wavelet transform on a transform coefficient set with a digital watermark embedded with a previous position marker apparatus. A discrete wavelet transform unit that generates and outputs a set of transform coefficient sets by discrete wavelet transform using oversampling from the input discrete image data with embedded digital watermarks with position markers,
A digital watermark code detection unit that determines a code string for decoding a digital watermark based on a reliability calculated from each of the transform coefficients included in the set of transform coefficient sets and a digital watermark code;
A position marker detection unit for detecting a position marker from the code string output from the digital watermark code detection unit;
A digital watermark decoding unit that calculates a positional deviation amount based on the detected position marker, determines the position of a code string to be decoded using the obtained positional deviation amount, and then decodes information embedded as a digital watermark An electronic watermark detection apparatus comprising: A moving image frame storage unit for storing the input moving image data by frame;
A discrete wavelet transform unit that outputs a set of transform coefficients by subjecting the extracted frame to discrete wavelet transform;
A code that generates a code string with a position marker by inserting a position marker consisting of a code string that can be distinguished from information embedded as a digital watermark into a code string obtained by encoding information to be embedded as an input digital watermark And position marker insertion part,
A digital watermark embedding unit that generates a position marker-added digital watermark embedded transform coefficient set from the transform coefficient set and the position marker-added code string;
An inverse discrete wavelet transform unit for generating discrete image data with embedded position-added digital watermark by performing inverse discrete wavelet transform on the transform coefficient set with embedded position-marked digital watermark; and
An electronic watermark embedding apparatus comprising: a moving image reconstructing unit that reconstructs moving image data with embedded digital watermarks with position markers from the discrete image data with embedded digital watermarks with position markers. A moving image frame storing unit that stores the input moving image data embedded with the digital watermark with the position marker for each frame;
A discrete wavelet transform unit that outputs a set of transform coefficient sets from the extracted frame by discrete wavelet transform using oversampling;
A digital watermark code detecting unit for calculating a reliability and a digital watermark code string from each of the transform coefficients included in the set of transform coefficient sets;
A reliability and code string storage unit that stores the reliability and code string output from the digital watermark code detection unit;
A decoding code determination unit that determines a code string for decoding a digital watermark based on the stored reliability and code string;
A position marker detection unit for detecting a position marker from the code string output from the decoding code determination unit;
A digital watermark decoding unit that calculates a positional deviation amount based on the detected position marker, determines the position of a code string to be decoded using the obtained positional deviation amount, and then decodes information embedded as a digital watermark An electronic watermark detection apparatus comprising: A moving image frame storing unit that stores the input moving image data embedded with the digital watermark with the position marker for each frame;
A frame distribution unit that switches a destination of a frame output from the frame storage unit;
A first discrete wavelet transform unit that outputs a set of transform coefficient sets from a frame output from the frame sorting unit by discrete wavelet transform using oversampling;
A second discrete wavelet transform unit that outputs a set of specific transform coefficient sets from a frame output from the frame sorting unit by a discrete wavelet according to a designated sampling method;
A digital watermark code detection unit that calculates a reliability and a digital watermark code sequence from each of the transform coefficients included in a set of transform coefficients output from the first discrete wavelet transform unit and the second discrete wavelet transform unit; ,
A reliability and code string storage unit that stores the reliability and code string output from the digital watermark code detection unit;
A digital watermark presence / absence determining unit that controls the frame distribution unit based on the accumulated reliability and code string;
A decoding code determination unit that determines a code string for decoding a digital watermark based on the stored reliability and code string;
A position marker detection unit for detecting a position marker from the code string output from the decoding code determination unit;
A digital watermark decoding unit that calculates a positional deviation amount based on the detected position marker, determines the position of a code string to be decoded using the obtained positional deviation amount, and then decodes information embedded as a digital watermark An electronic watermark detection apparatus comprising:

Images