KR100821349B1 - Method for generating digital watermark and detecting digital watermark - Google Patents

Abstract

According to the present invention, there is provided a method, comprising: a frame division step of dividing a digital signal into one or more frame signals; A wavelet transform step of wavelet transforming the frame signal into a predetermined number of band signal groups divided by frequency bands; A conversion target band signal group normalization step of statistically normalizing a band signal group set for conversion among a predetermined number of band signal groups; A transform target corresponding to wavelet reference coefficients of a low frequency band signal group having a magnitude greater than or equal to a threshold value among the normalized transform target band signal groups based on a threshold using a standard deviation and a predetermined coefficient of the low frequency band signal group among the band signal groups A predetermined wavelet coefficient of the band signal group is selected, and the selected wavelet coefficients of the band signal group to be converted are divided into two groups according to the sign of the low frequency reference coefficients; A watermarking step of inserting a watermark as each variance value of the two groups is modified; An inverse transform step of inverse normalizing a band-target signal group having a watermark embedded therein, inverse wavelet-converting a predetermined number of band-signal groups, and generating a plurality of frame signals; And a frame merging step in which each frame signal is merged and a digital signal is generated.

Digital audio, watermarks, wavelet transforms, frames, statistical methods

Description

Method for generating digital watermark and detecting digital watermark

1 is a system configuration diagram for realizing a digital watermark generation method according to an embodiment of the present invention;

2 is a flowchart of a digital watermark generation method according to an embodiment of the present invention;

3 is a schematic diagram illustrating a frame division step and a wavelet transform step of FIG. 2;

FIG. 4 is a schematic diagram illustrating a probability distribution change of each group in the watermarking step of FIG. 2.

<Description of the symbols for the main parts of the drawings>

10 ... digital signal 20 ... frame signal

30 ... band signal group 31 ... low frequency band signal group

100 ... frame divider 110 ... wavelet transform

120 ... Regularization 130 ... Division

140 Watermark insertion section 150 Normalizing section

160 ... wavelet inverse transform unit 170 ... frame merger

The present invention relates to a digital watermark generation method and a detection method, and more particularly, to a digital watermark generation method and detection method for inserting a watermark, which is the owner's ownership information, in a digital content to protect the copyright of the content. will be.

In general, digitized data is easily reproduced and retransmitted due to the development of multimedia data technology, and accordingly, the demand for copyright protection of producers is increasing.

Conventionally, as a technology for protecting such digitized data, an encryption technique has been proposed to control user access so that only an authenticated user can use the data, but after the data is decrypted by the authenticated user, Since it is not free from modification or duplication, there is a problem in that stable protection of copyright protection is difficult.

Unlike the encryption technology, according to the watermark generation method, the access of the user is not controlled, but as the copyright information of the producer is subordinated to the content by the insertion of the watermark, By extracting the inserted watermark, not only the copyright can be claimed but also the owner's ownership can be protected.

In this case, the digital signal of the content may be deteriorated due to signal distortion caused by the insertion of a watermark, so that the watermark is inaudible to the digital signal that cannot be detected by human hearing while minimizing the signal distortion as described above. Inaudibility should be guaranteed.

In addition, the watermark must have robustness so that the watermark is not damaged or modified at all in the external attack signals such as amplitude change, noise addition, filtering, resampling, MP3 compression, and file transfer.

In general, the low frequency band of a digital signal may be suitable as an area to insert a watermark due to its robustness which is relatively unaffected by an external attack signal, but the degree of distortion of the digital signal due to the watermark insertion is not good because of inaudibility. There is a problem that can be easily detected by the human hearing, the commercial value is lowered.

In the related art, a watermark generation method based on statistical characteristics of a digital signal capable of embedding a robust watermark relatively insensitive to the external attack signal has been proposed.

Among them, the Fourier transform of a digital signal adds or subtracts an arbitrary constant to two groups selected arbitrarily to change the average value. The discrete cosine transform modifies the transform coefficient in proportion to the standard deviation of the transform coefficient. Patchwork methods such as the scheme have been proposed.

However, the above method requires accurate location information of audio samples or conversion coefficients when watermarks are detected, and the detection rate of the watermarks increases as the location information of the digital signal is sensitively changed during time-varying attacks or MP3 compression among external attack signals. There is a problem that can be lowered and thus the guarantee of copyright protection is lowered.

On the other hand, there is also a watermark embedding method of modifying the average value of the two groups separated in the time domain according to the polarity of the low-pass filtered audio sample, the watermark is inserted into the low-frequency components to the external attack signal Toughness may be excellent, but there is an inefficient disadvantage of watermark inaudibility.

The present invention was created to solve the above-mentioned problems, and is based on a statistical method for digital signals, and has an inaudibility and robustness of an inserted watermark, thereby improving digital watermark generation and detection. The purpose is to provide a method.

Another object of the present invention is to provide a computer-readable recording medium having recorded thereon a program for executing the digital watermark generation method and detection method on a computer.

Other objects and advantages of the invention will be described below and will be appreciated by the embodiments of the invention. Furthermore, the objects and advantages of the present invention can be realized by means and combinations indicated in the claims.

The digital watermark generation method of the present invention for achieving the above object comprises a frame division step of dividing a digital signal into one or more frame signals; A wavelet transform step of converting the frame signal into a predetermined number of band signal groups divided by frequency bands by wavelet transforming the frame signal; A conversion target band signal group normalization step of statistically normalizing a band signal group set for conversion among the predetermined number of band signal groups; Wavelet reference coefficients of the low frequency band signal group having a magnitude greater than or equal to the threshold value among the normalized conversion target band signal group based on a threshold value using a standard deviation and a predetermined coefficient of the low frequency band signal group among the band signal group; A conversion target band signal group dividing step, wherein predetermined wavelet coefficients of a corresponding conversion target band signal group are selected, and the selected wavelet coefficients of the conversion target band signal group are divided into two groups according to the sign of the low frequency reference coefficients; A watermarking step of inserting a watermark as each variance value of the two groups is modified; An inverse transform step of inverse normalizing a conversion target band signal group into which a watermark is inserted, and inversely wavelet transforming the predetermined number of band signal groups to generate a plurality of frame signals; And a frame merging step in which the respective frame signals are merged and a digital signal is generated.

Further, the threshold value is a product of a predetermined coefficient α to which the selection area of the wavelet coefficients of the band signal group to be converted is adjusted and the standard deviation σ _c of the low frequency band signal group (| α × σ _c |). Can be made.

In the dividing step of the target band signal group, the wavelet reference coefficients c [of the low frequency band signal group having a magnitude greater than or equal to the threshold value | α × σ _c | are obtained by using Equations 4a and 4b. selecting predetermined wavelet coefficients đ [l] of the band band to be converted; And dividing the selected wavelet coefficients đ [l] of the band signal group to be converted into two groups, A and B, according to the sign (+ or −) of the low frequency reference coefficients c [l]. It may include.

A = {đ [l] | l ∈ Ω ⁺ }, Ω ⁺ = {l | c [l]> α × σ _c }

B = {đ [l] | l ∈ Ω -}, Ω - = {l | c [l] <-α × σ c}

In addition, a watermark of 1 bit may be inserted into each frame signal.

In addition, in the watermarking step, the dispersion value of one group may be reduced and the dispersion value of another group may be increased according to the type of watermark to be inserted.

Also, for the two groups A and B, when the inserted watermark ω is +1, the variance value of each group is equal to the standard deviation σ _A of the group A as shown in Equation 5a. the overall standard deviation, the standard deviation (σ _B) of large and the B group than (σ _đ) of normalized converted band wavelet coefficients of the signal group is converted to be smaller than the overall standard deviation (σ _đ), it included in each group The wavelet coefficient value of the target band signal group and the watermark embedding intensity (β) are respectively adjusted through adjustments,

(σ _A > (1 + β) × σ _đ ) & (σ _B <(1-β) × σ _đ ), ω = + 1

When the inserted watermark ω is -1, the variance value of each group is as shown in Equation 5b, where the standard deviation σ _A of the group A is smaller than the total standard deviation σ _đ and The standard deviation σ _B may be corrected by adjusting the wavelet coefficient value and the watermark embedding intensity β of the group of band signals to be included in each group so as to be larger than the total standard deviation σ _đ . .

(σ _A <(1-β) × σ _đ ) & (σ _B > (1 + β) × σ _đ ), ω = -1

In addition, the digital watermark detection method of the present invention includes two groups, A 'and B', from wavelet coefficients of the band signal group to be converted of the digital signal into which the watermark? Is generated. after the acquisition, wherein the insert through comparison of equation 6a and equation 6b said a 'standard deviation of the group (σ _a') and the B '(σ _B standard deviations of the group _") value using the watermark ( ω) may be detected.

ω '(detected watermark) = + 1, σ _A' > σ _{B '}

ω '(detected watermark) =-1, σ _A' <σ _{B '}

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Prior to this, terms or words used in the specification and claims should not be construed as having a conventional or dictionary meaning, and the inventors should properly explain the concept of terms in order to best explain their own invention. It should be interpreted as meanings and concepts corresponding to the technical idea of the present invention based on the principle of definition.

Therefore, the embodiments described in the specification and the drawings shown in the drawings are only the most preferred embodiment of the present invention and do not represent all of the technical idea of the present invention, various modifications that can be replaced at the time of the present application It should be understood that there may be equivalents and variations.

The digital watermark generation method of the present invention relates to a watermark generation method based on statistical characteristics of digital signals constituting digital content.

1 is a system configuration diagram for realizing a digital watermark generation method according to an embodiment of the present invention, FIG. 2 is a flowchart of a digital watermark generation method according to an embodiment of the present invention, and FIG. 4 is a schematic diagram illustrating a frame division step and a wavelet transform step, and FIG. 4 is a schematic diagram showing a change in probability distribution of each group in the watermarking step of FIG. 2.

Hereinafter, the digital watermark generation method of the present invention will be described in detail with reference to FIGS. 1 to 4.

First, as illustrated in FIG. 3, the digital signal 10 is divided into one or more frame signals 20 in the frame dividing unit 100 (S200).

Each of the frame signals 20 may mean a split signal according to the temporal flow of the digital signal 10.

Here, the digital signal 10 should be understood as a signal encompassing an audio signal, a video signal, and an image signal that can be represented on a time domain.

Each of these frame signals 20 may include a watermark of 1 bit.

After the frame division step S200, as shown in FIG. 3, each of the frame signals 20 is wavelet transformed by the wavelet transform unit 110 to be divided into frequency bands of a predetermined number of band signal groups 30. Is converted (S210).

Here, each of the band signal groups divided by a predetermined number has a plurality of wavelet coefficients by the wavelet transform.

The predetermined number may be at least two or more so that the band signal group 30 is converted into at least two or more frequency bands.

Referring to FIG. 3, a band signal group corresponding to a high frequency band is indicated from the left side to the right side of the predetermined number of band signal groups 30, and the band signal group 31 located on the left is a low frequency band signal. This group will be referred to as 31 or c [n].

After the wavelet transform step (S210), the band signal group set as a conversion target among the predetermined number of band signal groups 30 is statistically normalized by the normalization unit 120 (S220).

Here, the band signal set as the conversion target, that is, the conversion target band signal is selected from one of the band signal groups 32 other than the low frequency band signal group 31 among the predetermined number of band signal groups 30 of FIG. 3. It may be a band signal group.

That is, one of the band signal groups 32 of the band signal groups 32 corresponding to the high frequency band relative to the low frequency band signal group 31 means a potential conversion target band signal group that can be set as the conversion target. The set conversion target band signal group substantially corresponds to a portion where a watermark can be inserted.

When the watermark is inserted in the low frequency band, the distortion of the digital signal due to the deformation caused by the watermark insertion can be easily detected by the human auditory. In the present invention, the conversion target band signal corresponding to the high frequency band is relatively higher than the low frequency band. By embedding a watermark in the group, it is possible to enhance the inaudibility of the watermark.

The conversion target band signal group d [n] may be changed into a normalized conversion target band signal group đ [n] having a wavelet coefficient value in the range [-1,1] through a statistical normalization process. have.

After the normalization step (S220), based on the threshold value using the standard deviation (σ _c ) and the predetermined coefficient (α) of the low frequency band signal group 31 of the predetermined number of band signal group 30, the dividing unit The predetermined wavelet coefficients đ [l] are selected from the normalized conversion target band signal group đ [n] through 130, and the selected predetermined wavelet coefficients đ [l] are divided into two groups. It is divided (S230).

The threshold value is a predetermined coefficient α of which a selection area of wavelet coefficients of the normalized conversion target band signal group đ [n] is adjusted, and a standard piece of the low frequency band signal group c [n]. It can be made of the product of the difference σ _c '(| α × σ _c |).

Here, the standard deviation σ _c means a standard deviation of wavelet coefficient values constituting the low frequency band signal group c [n].

Hereinafter, the step (S230) of dividing the conversion target band signal group based on the threshold value and the equations 7a to 7b will be described in more detail.

Among the normalized conversion target band signal group đ [n], the wavelet reference coefficients of the low frequency band signal group having a magnitude greater than or equal to the threshold value α | σ _c | 'Predetermined Wavelet Coefficients of the Band Signal Group to be Converted' (đ [l]) is selected.

The selected wavelet coefficients đ [l] of the band signal group to be converted may be divided into two groups A and B according to the sign (+ or −) of the low frequency reference coefficients c [l]. have.

A = {đ [l] | l ∈ Ω ⁺ }, Ω ⁺ = {l | c [l]> α × σ _c }

B = {đ [l] | l ∈ Ω -}, Ω - = {l | c [l] <-α × σ c}

When the values of the low frequency reference coefficients c [l] are large, the selected wavelet coefficients đ [l] of the band signal group to be converted having a coefficient corresponding to the low frequency reference coefficients c [l] Since the size also has a large value corresponding thereto, when the watermark is inserted using the selected wavelet coefficients đ [l] later, the robustness of the inserted watermark may be increased.

On the other hand, when the value of the standard deviation (σ _c ) or the predetermined number (α) of the low-frequency band signal group is large, the threshold is also proportionally increased. Accordingly, low-frequency reference coefficients having a magnitude greater than or equal to the increased threshold value ( The number of selected samples of c [l]) is reduced.

In this case, since the number of samples of the wavelet coefficients đ [l] of the band band signal group to be selected corresponding to the low frequency reference coefficients c [l] is reduced, the wavelet coefficients of the band band signal group to be converted are reduced. Since a watermark can be inserted with a smaller amount of sample correction, the distortion of the digital signal can be minimized, thereby improving the inaudibility and audio quality of the watermark.

On the other hand, since the standard deviation σ _c of the low frequency band signal group constituting the threshold value is a fixed value determined from the digital signal, the threshold value is adjusted by changing the predetermined number α. Toughness and inaudibility can also be adjusted.

Next, as illustrated in FIG. 4, the watermark inserting unit 140 inserts a watermark as the respective dispersion values of the A group and the B group corresponding to the two groups are modified (S240).

As shown, the probability distributions (a) for the two groups A and B before the variance values are corrected have similar distributions, that is, the Laplacian distribution, while after the variance values have been modified (b) or Like the probability distribution in (c), it can be seen that the variance of one group is modified to be larger or smaller than the variance of another group. This degree of dispersion can be seen through the arrows at the bottom of the graph as well as each graph curve of FIG. 4.

In this watermarking step (S240), according to the type of watermark to be inserted, that is, the watermark is +1 or -1, the dispersion value of the group A (or group B) is increased and the group B (or group A) is increased. The variance of can be modified to be reduced.

Here, since the variance value has a homology with the square of the standard deviation, the variance value of each group can be corrected by adjusting the standard deviation for each group.

Referring to Equation 8a, when the inserted watermark ω is +1, the variance value of each group includes wavelet coefficients of the normalized conversion target band signal group in which the standard deviation σ _A of the group _A is normalized. The wavelet coefficient value of the target band signal group included in each group is greater than the total standard deviation (σ _đ ) for, and the standard deviation (σ _B ) of the group _B is smaller than the total standard deviation (σ _đ ). đ [l]) and the watermark embedding strength β can be modified.

(σ _A > (1 + β) × σ _đ ) & (σ _B <(1-β) × σ _đ ), ω = + 1

On the contrary, when the inserted watermark ω is -1, the variance value of each group is as shown in Equation 8b, and the standard deviation σ _A of the group _A is smaller than the total standard deviation σ _đ . The standard deviation (σ _B ) of group _B is larger than the total standard deviation (σ _đ ) so that the wavelet coefficient value (đ [l]) and the watermark embedding strength (β) of the band signal group to be converted in each group are included. Each can be modified by adjusting.

(σ _A <(1-β) × σ _đ ) & (σ _B > (1 + β) × σ _đ ), ω = -1

Here, since the standard deviation σ _A or σ _B of each group is determined by wavelet coefficient values đ [l] of the band band signal group to be included in each group, the wavelet coefficient value đ According to the correction of [l]), the standard deviation (σ _A or σ _B ) of each group can be corrected. Of course, the variance value for each group is also corrected according to the correction of the standard deviation σ _A or σ _B of each group.

The wavelet coefficient value đ [l] included in each group, which is adjusted at the time of modifying the variance value for each of the groups, may be decreased or increased by adjusting the k value of Equation 9 below.

đ '[l] = sign (đ [l]) × | đ [l] | ^k , 0≤k <∞

Referring to Equation 9, the variance value of each group may be increased when k is in the range of [0,1] and decreased when k is in the range of [1, ∞].

That is, as the k value is adjusted, the wavelet coefficient value đ [l] included in each group is changed, and the standard deviation or variance value is decreased or decreased by each wavelet coefficient value. Can be modified to increase.

On the other hand, as the watermark embedding strength β increases, the standard deviation σ _A or σ _B of each group is corrected to a larger or smaller value around the total standard deviation σ _đ . The gap between the standard deviation σ _{A of the} group and the standard deviation σ _B of the B group increases, and thus the robustness of the inserted watermark may increase.

The watermark embedding strength β may be inserted to increase the robustness of the watermark to be inserted. However, since the embedding strength increases, the original digital signal may be distorted as the embedding strength increases, thereby increasing the toughness of the watermark. At the same time, it is desirable to adjust the range within which the distortion of the digital signal can be minimized.

After the watermarking step (S240), the conversion target band signal group into which the watermark is inserted is denormalized by the denormalization unit 150, and the predetermined number of band signal groups are deformed by the wavelet inverse conversion unit 160. The plurality of frame signals are converted and generated again (S250).

Of course, the subject of the denormalization corresponds to all of the band band signals to be normalized in the normalization step (S220), so that the denormalization step (S250) is divided into A or B groups in the band band splitting step (S230). All the conversion target band signal groups including wavelet coefficients of the unselected conversion target band signal are denormalized.

In addition, the object of the inverse wavelet transform corresponds to the entire number of band signal groups including the denormalized transform target band signal.

Next, the frame merger 170 merges the respective frame signals and generates a digital signal in which a watermark is inserted (S260).

By the above method, the copyright of the producer is protected as the watermarking is inserted in the hidden state in the original digital signal.

On the other hand, the watermark generated by the digital watermark generation method according to the present invention as described above is as follows.

According to the watermark detection method, after two groups of A 'and B' are obtained from wavelet coefficients of a band signal group to be converted of a digital signal into which a watermark? using 10a) and (10b wherein a 'standard deviation (σ _a of group _") and the B' standard deviation (σ _B of the group _') watermarked the insertion by comparison of the value (ω) can be detected have.

ω '(detected watermark) = + 1, σ _A' > σ _{B '}

ω '(detected watermark) =-1, σ _A' <σ _{B '}

That is, when the standard deviation σ _{A '} of the A' group is greater than the standard deviation σ _{B '} of the B' group so as to correspond to the watermarking step S240, a watermark of +1 is detected. On the contrary, when the standard deviation σ _{A '} of the A' group is smaller than the standard deviation σ _{B '} of the B' group, a watermark of −1 may be detected.

In this method, since the inserted watermark can be detected by simply comparing the difference between the variance values from the two groups of the band signal group to be converted, the watermark can be easily detected and the watermark can be detected without the original signal. There is an advantage that the blind technique can be realized.

As described above, by the method of detecting the watermark from the digital signal in which the watermark is inserted, not only the copyright of the modified or copied digital content can be claimed, but also the owner's ownership can be protected therefrom.

Hereinafter, the effects on the inaudibility and robustness in inserting the watermark of the present invention through simulation data by the digital watermark generation method of the present invention will be described.

The digital signal used in the simulation is 44.1kHz, 16bits mono classical data. In this experiment, for convenience of explanation, each frame 22 will be described only in the case where the total band signal group is divided into five bands.

Table 1 below shows frequency ranges for the decomposed band signal groups.

TABLE 1

1st band signal group (1st-band) 2nd band signal group (2nd-band) 3rd band signal group (3rd-band) 4th band signal group (4th-band) 5th band signal group (5th-band) Frequency range (kHz)  0.0-1.3  1.3-2.7  2.7-5.5  5.5-11.0  11.0-22.0

In Table 1, the first band signal group corresponds to the low frequency band signal group 31 of FIG. 3, and the second band signal group to the fifth band signal group correspond to the conversion target band signal group 32.

According to the watermark embedding method of the present invention, the watermark is inserted into one band signal group selected from the second band signal group and the fifth band signal group, and no deformation is applied to the remaining band signal group that is not selected. Do not.

Table 2 shows the watermark embedding strength applied to each of the band signal group to be converted. In the simulation, the predetermined coefficient α is fixed to 1, and the threshold value | α × σ _c | Only the standard deviation (σ _c ) value of the low frequency signal group was used.

TABLE 2

 Band signal group for conversion 2nd band signal group (2nd-band) 3rd band signal group (3rd-band) 4th band signal group (4th-band) 5th band signal group (5th-band) Watermark embedding strength (β)  0.3  0.45  0.6  0.75

Since the high frequency band is less sensitive than the low frequency band due to the increase of the watermark embedding strength β, in this experiment, the embedding intensity β of the watermark increases as the high frequency band increases. Frequency-specific adjustment of the watermark embedding intensity (β) is only one embodiment of the present invention, of course, it can be modified in various ways.

On the other hand, the inaudibility of the watermark can be determined by measuring the signal to noise ratio (SNR).

As the watermark is inserted, the inserted watermark may act as noise with respect to the original digital signal. When the SNR value is large, the noisemark is weak compared to the digital signal, indicating that the inaudible and sound quality are excellent.

On the other hand, external attack factors for the robustness test of the inserted watermark include 2: 1 resampling (2: 1 down sampling and up sampling through bi-linear interpolation) and band pass filtering (0.1 to 6 kHz). ), Echo Embedding (amplitude: 0.5, delay: 100msec), equalizer (Equalization, -6 ~ 6dB), MP3 compression (128kbps, 64kbps), and adding White Noise.

The robustness of the watermark may be determined by measuring a distance (DR).

The DR is a value obtained by dividing the number of inserted watermark bits by the number of accurately detected watermark bits (number of detected watermark bits / number of inserted watermark bits), and the watermark extracted as the DR value increases. Since the number of bits is large, it indicates that the watermark is robust against external attack elements.

Table 3 below shows measurement results of SNR and DR values when the external attack element is applied to the corresponding band signal group with the watermark inserted.

TABLE 3

  External attack element 2nd band signal group (2nd-band) 3rd band signal group (3rd-band) 4th band signal group (4th-band) 5th band signal group (5th-band) SNR DR SNR DR SNR DR SNR DR Before external attack 23.37 1.00 24.37 1.00 23.10 1.00 35.00 1.00 2: 1 resampling 21.69 1.00 22.35 1.00 24.32 1.00 26.55 0.48 Band Pass Filtering 5.79 0.94 5.81 0.91 5.85 0.43 5.86 0.64 Echo Insert 4.69 0.91 4.69 0.99 4.73 1.00 4.75 1.00 Equalizer 6.79 0.99 6.78 1.00 6.78 1.00 6.85 1.00  MP3 compression 128 kbps 23.00 1.00 23.90 1.00 26.90 1.00 30.17 1.00 64 kbps 18.60 1.00 18.87 1.00 19.11 1.00 19.83 0.99  White noise added 100 22.05 1.00 22.77 0.99 24.98 0.99 27.11 0.99 900 8.65 0.97 8.68 0.96 8.74 0.89 8.79 0.68 1700 3.23 0.89 3.24 0.86 3.26 0.73 3.27 0.56 medium 13.79 0.97 14.15 0.97 15.28 0.90 16.82 0.83

'Before external attack' in Table 3 means that external attack element is not applied while the watermark is inserted into the digital signal, and the SNR value of each band signal group shows excellent characteristics of 23 or more. It can be seen that the inaudibility is also excellent.

In addition, since the external attack element does not penetrate the digital signal before the external attack, the ratio of 1 in which the number of bits of the inserted watermark and the detected watermark is equal to the DR value of each band signal group can be maintained.

On the other hand, when various external attacks are applied, the SNR value is reduced when bandpass filtering, echo insertion, equalizer, and white noise are added, but shows excellent characteristics with little change during 2: 1 resampling and MP3 compression.

On the other hand, when various external attacks are applied, the average of the DR value represents a value of 0.83 or more for the entire band signal group, which means that 83% or more of the inserted watermarks can be detected, indicating that the watermark has excellent robustness. Can be.

Therefore, according to the digital watermark generation method of the present invention, when illegally modifying or duplicating a digital signal, damage or deformation rate of the inserted watermark may be small, thereby ensuring the protection of copyright protection.

Here, the average value of the DR increases with the low frequency band signal group due to the frequency inherent characteristics of the low frequency band superior to that of the low frequency band.

The present invention can be embodied as computer readable codes on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data is stored as a medium that can be read by a computer system.

Examples of computer-readable recording media include ROM, RAM, CO-ROM, magnetic tape, floppy disks, optical data storage devices, and the like, which are also implemented in the form of carrier waves (for example, transmission over the Internet). Include. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

As described above, although the present invention has been described by way of limited embodiments and drawings, the present invention is not limited thereto and is described by the person of ordinary skill in the art to which the present invention pertains. Various modifications and variations are possible without departing from the scope of the appended claims.

The digital watermark generation method according to the present invention provides the following effects.

First, the watermark may be improved by inserting a watermark by modifying a variance value of a band of a band band signal group corresponding to a high frequency band relative to a low frequency band.

Second, as the watermark is inserted using wavelet coefficients of the band band group to be converted corresponding to the low frequency reference coefficients higher than or equal to a predetermined threshold, the robustness of the inserted watermark may be improved.

Third, a blind technique capable of easily detecting an inserted watermark and detecting a watermark without an original signal can be realized by simply comparing a difference between dispersion values from two groups of the band signal group to be converted. Can be.

Claims (8)

A frame division step of dividing the digital signal into one or more frame signals; A wavelet transform step of converting the frame signal into a predetermined number of band signal groups divided by frequency bands by wavelet transforming the frame signal; A conversion target band signal group normalization step of statistically normalizing a band signal group set for conversion among the predetermined number of band signal groups; Wavelet reference coefficients of the low frequency band signal group having a magnitude greater than or equal to the threshold value among the normalized conversion target band signal group based on a threshold value using a standard deviation and a predetermined coefficient of the low frequency band signal group among the band signal group; A conversion target band signal group dividing step, wherein predetermined wavelet coefficients of a corresponding conversion target band signal group are selected, and the selected wavelet coefficients of the conversion target band signal group are divided into two groups according to the sign of the low frequency reference coefficients; A watermarking step of inserting a watermark as each variance value of the two groups is modified; An inverse transform step of inverse normalizing a conversion target band signal group into which a watermark is inserted, and inversely wavelet transforming the predetermined number of band signal groups to generate a plurality of frame signals; And A frame merging step of merging each frame signal and generating a digital signal, In the watermarking step, And the dispersion value of one group is decreased and the dispersion value of another group is increased according to the type of watermark to be inserted. The method of claim 1, wherein the threshold value, And a product (| α × σ _c |) of a predetermined coefficient α to which the selection area of the wavelet coefficients of the band band signal group to be converted is adjusted and the standard deviation σ _c of the low frequency band signal group. How to create a watermark. 3. The method of claim 2, wherein the dividing of the band signal group to be converted comprises: Using Equations 1a and 1b, The predetermined wavelet coefficients đ [l] of the target band signal group to be converted corresponding to the wavelet reference coefficients c [l] of the low frequency band signal group having a magnitude greater than or equal to the threshold value | α × σ _c | Is selected; And The selected wavelet coefficients đ [l] of the band signal group to be converted may be divided into two groups A and B according to the sign (+ or −) of the low frequency reference coefficients c [l]. Digital watermark generation method, characterized in that. A = {đ [l] | l ∈ Ω ⁺ }, Ω ⁺ = {l | c [l]> α × σ _c } B = {đ [l] | l ∈ Ω -}, Ω - = {l | c [l] <-α × σ c} The method of claim 1, And a 1-bit watermark is inserted into each of the frame signals. delete The method of claim 5, For the two groups A group and B group, When the inserted watermark (ω) is +1, the variance of each group is As shown in Equation 2a, the standard deviation σ _A of the group A is greater than the total standard deviation σ _đ of wavelet coefficients of the normalized conversion target band signal group, and the standard deviation σ _B of the group _B is By adjusting the wavelet coefficient value and the watermark embedding intensity (β) of the group of band signals to be converted included in each group so as to be less than the total standard deviation (σ _đ ), (σ _A > (1 + β) × σ _đ ) & (σ _B <(1-β) × σ _đ ), ω = + 1 When the inserted watermark (ω) is -1, the variance of each group is As shown in Equation 2b, the standard deviation (σ _A ) of the group A is smaller than the total standard deviation (σ _đ ) and the standard deviation (σ _B ) of the group _B is larger than the total standard deviation (σ _đ ), A digital watermark generation method characterized by modifying the wavelet coefficient value and the watermark embedding strength (β) of the band signal group to be included in each group, respectively. (σ _A <(1-β) × σ _đ ) & (σ _B > (1 + β) × σ _đ ), ω = -1 A method for detecting a digital watermark generated by the digital watermark generation method according to claim 1, comprising: After the two groups A 'and B' are obtained from wavelet coefficients of the band signal group of the digital signal into which the watermark ω is inserted, equations 3a and 3b are used to calculate the group A '. And a watermark detection step of detecting the inserted watermark (ω) by comparing a standard deviation (σ _{A '} ) and a standard deviation (σ _B' ) value of the B 'group. Detection method. ω '(detected watermark) = + 1, σ _A' > σ _{B '} ω '(detected watermark) =-1, σ _A' <σ _{B '} A computer-readable recording medium having recorded thereon a program for executing the method according to any one of claims 1 to 4, 6 and 7.

Images