WO2017188500A1

WO2017188500A1 - Watermark embedding and extracting device, and watermark embedding and extracting method thereof

Info

Publication number: WO2017188500A1
Application number: PCT/KR2016/005770
Authority: WO
Inventors: 이승룡; 티엔현더
Original assignee: 경희대학교 산학협력단
Priority date: 2016-04-29
Filing date: 2016-05-31
Publication date: 2017-11-02

Abstract

A watermark embedding method enables a watermark code comprising binary digits to be embedded in an input image having RGB channels, and comprises the steps of: wavelet-transforming an input image for each channel; extracting a plurality of watermark embedding areas from the wavelet-transformed input image; selecting a watermark embedding channel by using the bit value of the watermarking image and the differences in coefficients among the plurality of watermark embedding areas; and embedding a watermark code in the plurality of watermark embedding areas of the selected channel.

Description

【Specification】

[Name of invention]

Watermark embedding and extracting device, and its watermark embedding and extraction method

The present invention relates to a watermark embedding and extracting apparatus, and a watermark embedding and extracting method using the same.

Background technology

Recently, with the development of computers and networks, intellectual property problems due to illegal copying or distribution of digital contents such as digital video, audio, and video have emerged.

Illegal copying and distribution of digital content can lead to loss of author and creative motivation and serious economic loss. Therefore, copyright protection technology is required to prevent illegal copying without the consent of the owner and the copyright holder and to effectively protect the ownership.

With such copyright protection technology, digital watermarking has been focused on inserting ownership information into digital content itself to prevent illegal copying or unauthorized distribution and effectively protecting copyright. The digital watermarking technique is a technology for protecting the intellectual property rights and copyrights and providing a basis for claiming by inserting watermarks into the data of personal information of the owner or distributor of digital content.

Digital watermarking is a must-have requirement, which means that watermarks are transparently inserted to make the original media invisible (visually and audibly) invisible (invi s ibi li ty). Robustness in which an inserted watermark should be detected for an attack, and security that prevents the detection or removal of the presence of a watermark even if the embedding and detection algorithm of the watermark are known.

In particular, the two conditions of invisibility and robustness are higher depending on the strength of the watermark, so it should be considered in designing the algorithm.

[Detailed Description of the Invention] [Technical problem]

An object of the present invention is to provide a watermark embedding apparatus and method thereof which selects an optimal color channel for embedding a watermark and inserts a watermark only in the selected color channel to satisfy both of invisibility and robustness.

[Technical Solution]

A method for embedding a watermark in an input image by a watermark embedding apparatus according to an embodiment of the present invention, the method comprising: returning the input image to a wavelet for each color channel; and a plurality of watermarks in the wavelet-converted input image Extracting an embedding region, selecting a color channel into which the watermark is to be inserted using a bit value of the watermark image, and a coefficient difference between the plurality of watermark embedding regions, and extracting from the selected color channel And inserting a watermark code composed of binary numbers into the plurality of watermark embedding areas.

The plurality of watermark embedding regions includes a first watermark embedding region and a second watermark embedding region, and selecting a channel to embed the watermark comprises: the first watermark embedding region coefficient, and the first watermark embedding region; Calculating a difference between the coefficients of the watermark embedding region, determining a quantization threshold using the coefficient difference of the watermark embedding region, the watermark code data value, and the watermark embedding region And selecting the channel having the smallest difference between the coefficient difference and the quantization threshold as a watermark embedding channel.

The step of calculating a difference between the first watermark embedding region coefficient and the second watermark embedding region coefficient may include ascending the first watermark embedding region coefficient and the second watermark embedding region coefficient. You can sort by order and then calculate the sorted coefficient difference.

The setting of the quantization threshold may include setting a first quantization threshold and setting a second quantization threshold having a value greater than the first quantization threshold.

The selecting of the watermark embedding channel may be performed by determining the coefficient difference between the first quantization threshold and the aligned coefficient difference in an area where data of the watermark code is zero. The channel having the smallest difference between the second quantization threshold and the aligned coefficient difference in the region where the data of the watermark code is 1 is selected as the color channel into which the watermark is to be inserted. The extracting of the plurality of watermark embedding regions may extract the LH region and the HL region as the watermark embedding region among the subbands in which the input image is wavelet transformed for each color channel.

The step of wavelet converting the input image for each color channel may be determined in consideration of the length and insertion strength of the watermark code and the degree of image quality deterioration due to the insertion of the watermark code.

An apparatus for embedding a watermark in an input image according to an embodiment of the present invention includes a wavelet transform unit for wavelet transforming the input image for each color channel, and a water extracting region for extracting a plurality of watermark embedding regions from the wavelet transformed input image. A mark selection unit extracting unit, a channel selection unit which selects a color channel into which the watermark is to be inserted using a bit value of the watermark image, and a coefficient difference between the watermark embedding regions, and the plurality of channels of the selected channel And a watermark embedding unit for embedding the watermark in a watermark embedding region of the unit.

The plurality of watermark embedding regions may include a first watermark embedding region and a second watermark embedding region, and the channel selector may include a difference between the first watermark embedding region coefficient and the second watermark embedding region coefficient. And a quantization threshold determination unit for determining a quantization threshold using the coefficient difference of the watermark embedding region and the watermark code data value.

The channel selector may select a channel having the smallest difference between the coefficient difference of the watermark embedding region and the quantization threshold as a watermark embedding channel. The channel selector may arrange the first watermark embedding region coefficient and the second watermark embedding region coefficient in ascending order and then calculate the sorted coefficient difference.

The quantization threshold may include a first quantization threshold and a second quantization threshold having a value greater than the first quantization threshold. The channel selector is a difference between the first quantization threshold and the aligned coefficient difference in the region where the data of the watermark code is 0, and the second quantization threshold in the region where the data of the watermark code is 1. And the channel having the smallest difference between the sorted coefficients may be selected as the color channel into which the watermark is to be inserted.

The watermark embedding region extracting unit may extract the LH region and the HL region all watermark embedding region from the subbands obtained by wavelet transforming the input image for each color channel.

The wavelet converter may determine the number of stages in consideration of the length and insertion strength of the watermark code and the degree of image quality deterioration due to the insertion of the watermark code.

A method of extracting a watermark from a watermark embedding image by the apparatus for extracting watermarks according to an embodiment of the present invention, the method comprising: obtaining a plurality of watermark embedding area wavelet coefficients by performing discrete wavelet transform of the watermark embedding image; Blocking a plurality of watermark embedding region coefficients, identifying a block containing the watermark code using an associative key, calculating a plurality of watermark embedding region coefficients, Calculating an adaptive reference value to a value that minimizes dispersion of the plurality of watermark embedding region coefficient differences, and extracting the watermark code using the reference value and the plurality of watermark embedding region coefficient differences; . The method may further include calculating an original difference coefficient using an average value of the plurality of watermark embedding region coefficient differences.

The calculating of the red-eye reference value may include calculating the red-eye reference value as a value for minimizing the variance of the original difference coefficient.

The watermark code value is extracted as 1 when the coefficient difference of the plurality of watermark embedding areas is greater than or equal to the watermark embedding value by comparing the red-eye reference value with the coefficient difference between the plurality of watermark embedding areas. When the coefficient difference of the mark insertion region is smaller than the adaptation reference value, the watermark code value may be extracted as zero.

【Effects of the Invention】 According to the equation derived by the present invention, it is possible to select a color channel into which a watermark is to be inserted, thereby producing a watermark embedding image capable of satisfying both invisibility and robustness at the same time.

[Brief Description of Drawings]

1 is a block diagram schematically showing an embodiment of a configuration of a watermark embedding apparatus according to an embodiment of the present invention.

2 is a result of applying discrete wavelet transform (DWT) to an image to which a watermark is to be inserted.

3 is a flowchart of a method for producing a watermark embedded image according to the present invention. 4 is a flow chart illustrating in detail an embodiment of the optimal color channel selection step of FIG.

5 is a diagram illustrating an example of a watermark image.

FIG. 6 is a flowchart illustrating an exemplary embodiment of inserting the watermark code into a plurality of watermark embedding regions of the selected channel of FIG. 3.

7 is a flowchart of extracting a watermark code from a watermark-embedded image by the apparatus for extracting watermarks according to an embodiment of the present invention.

[Specific contents to carry out invention]

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention. As those skilled in the art would realize, the described embodiments may be modified in various different ways, all without departing from the spirit or scope of the present invention. The drawings and description are to be regarded as illustrative in nature and not restrictive. Like reference numerals designate like elements throughout the specification. In addition, in the case of the well-known technology, the detailed description thereof will be omitted.

The embodiments described in the specification and the configuration shown in the drawings are only one of the most preferred embodiments of the present invention and do not represent all of the technical idea of the present invention, various equivalents that may be substituted for them at the time of the present application It should be understood that there may be variations and variations.

Hereinafter, an image according to an embodiment of the present invention will be described with reference to the drawings. A watermark embedding apparatus and a watermark embedding method using the same will be described.

Referring to FIG. 1, the watermark embedding apparatus 100 according to an embodiment of the present invention includes a watermark embedding region extracting unit 110, a channel selecting unit 120, a watermark inserting unit 130, and a watermark embedding. And an image generator 140.

The watermark insertion region extraction unit 110 applies a discrete wavelet transform (DWT) to the image 10 to which the watermark is to be inserted and extracts a band to insert the watermark.

The wavelet transform used in image processing refers to a conversion process that repeats a process of decomposing an image into several different frequency bands using a pair of low pass and high pass filters.

In general, four different subbands are generated by decomposing an image signal, which is a two-dimensional signal, using a low pass filter and a high pass filter in the vertical and horizontal directions, respectively.

2 is a result of applying discrete wavelet transform (DWT) to an image to which a watermark is to be inserted. Referring to FIG. 2, it can be seen that four subbands of LL1, HL1, LH1, and HH1 are generated by applying discrete wavelet transform to an image signal.

LL1 is a band created by applying a low pass filter in the horizontal and vertical directions to the original image, and HL1 is a band created by applying a high pass filter in the vertical direction to the original image. In addition, LH1 is a band generated by applying a high pass filter in the horizontal direction to the original image and includes an error component of the horizontal frequency. HH1 is a band created by applying a high pass filter in the horizontal and vertical directions to the original image. to be.

That is, the lowest low frequency band LL1 is located at the upper left, and the highest high frequency band HH1 is located at the lower right.

For natural video, the energy of the signal in the horizontal and vertical low frequency bands (LL1) As it concentrates, most of the information of the image is contained in the low frequency band LL1. As shown in the right image 20 of FIG. 2, it can be considered as another new image.

That is, the wavelet decomposition can be applied once again to the low frequency band LL1, and as a result, the bands of the subbands LL2, HL2, LH2, and HH2 are generated.

Since the number of the wavelet transform stages determines the size of the DC region to which the watermark is to be inserted, it is possible to determine an appropriate number of wavelet transform stages in which image quality degradation due to the watermark insertion does not occur. In general, when performing n-stage wavelet transform on a PXQ-sized image, the size of the watermark insertion region may be determined as. At this time. As described above, the size of the watermark embedding region is preferably determined in consideration of the length and insertion strength of the watermark data string, and the degree of image quality deterioration due to the insertion of the watermark data string. Explain it as being done.

Accordingly, as shown in FIG. 2, 13 sub-bands (1, HLl, LHl, HH2, HL2, LH2, HH3, HL3, LH3, HH4, HL4, LH4, and LL4) are formed. do. Since the HH4 band is a low energy band, it is easy to be removed by an external attack such as an image processing compression process such as JPEG, etc. (i.e. low robustness), and the LL4 band is weak to noise. Determine the horizontal high frequency vertical low frequency band (HL) and the horizontal low frequency vertical high frequency band (LH).

Referring again to FIG. 1, the channel selector 120 selects a channel to insert a watermark in an image signal including three channels (R, G, and B channels).

The channel selector 120 selects an optimal channel to insert the watermark from three channels by using the bit value of the watermark code 20 having a binary value and the coefficient difference between the watermark embedding regions. At this time, the process of selecting the channel will be described in detail in the description of FIG. The watermark embedding unit 130 inserts the watermark by changing the coefficients of the watermark embedding regions LH4 and HL4 extracted by the watermark embedding region extracting unit 110 in the channel selected by the channel selector 120. In this case, a process of changing the coefficient of the watermark embedding region of the selected channel will be described in detail later with reference to FIG. 5.

The watermark embedding image generating unit 140 applies an inverse discrete wavelet transform (IDWT) to the watermark embedding result of the watermark embedding unit 130 to generate the watermark-embedded image 30.

Due to the configuration of the watermark embedding region extracting unit 110, the channel selecting unit 120, the watermark embedding unit 130, and the watermark embedding image generating unit 140, an image signal is obtained by using the coefficient difference of the embedding region. A watermark is inserted by selecting an optimal color channel for embedding the R, G, B channel augmentation watermark and changing the coefficients of the watermark embedding areas LH4 and HL4 in the selected color channel. As a result, a watermark embedded image capable of satisfying both invisibility and robustness can be produced. 3 is a flowchart of a method for producing a watermark embedded image according to the present invention. Referring to FIG. 3, first, a band to insert a watermark is extracted from an image 10 to which a watermark is to be inserted (S110).

In order to extract the watermark insertion regions HL and LH, discrete wavelet transform is applied to the original image 10 to which the watermark is to be inserted, and the image 10 to which the watermark is to be inserted is classified into four subbands. The discrete wavelet transform is applied to the selected low frequency band LL again. In this case, the number of transform stages may be determined in consideration of the length and insertion strength of the watermark data string and the degree of image quality deterioration due to the insertion of the watermark data string.

In this embodiment, four-stage wavelet transform is performed and thirteen sub-bands (HH1, HLl, LHl, HH2, HL2, LH2, HH3, HL3, LH3, HH4, HL4, LH4,

LL4) is generated, and a pair of intermediate frequency regions HL4 and LH4 having high robustness and noise resistance are determined as the watermark embedding region.

The tenacity, security, and A formula for selecting an optimal color channel capable of maintaining invisibility is derived (S120).

The optimal color channel can be selected using the coefficient difference of the watermark embedding region band and the bit value of the watermark code 20.

The watermark code is represented by a binary string value of 0 or 1, and may be generated by using a predetermined image or sign set by a user or by using a random function.

The watermark is embedded in the video signal by changing the coefficient of the watermark embedding region in the selected channel (S130).

Finally, an inverse discrete wavelet transform may be applied to an image signal having the watermark code added thereto, thereby generating an image having a watermark embedded therein (S140). 4 is a flow chart illustrating in detail an embodiment of the optimal color channel selection step of FIG.

First, the frequency band coefficients (coefficientK C _LHi , C _H¾ ) of the watermark embedding region are extracted from the wavelet transformed image and blocked for each channel and for each pixel (S210).

The difference between the regions where the watermark is to be inserted, that is, the coefficient between the first watermark embedding region LH4 and the second watermark embedding region HL4, is calculated as shown in Equation 1 below (S220).

[Equation 1]

C _L¾fc denotes the first watermark insertion region coefficient of the i th wavelet block of the k th channel (R, G, B), and C _{H k} is the i th wavelet block of the k th channel (R, G, B). The second watermark insertion area coefficient.

In the quantization step, according to the watermark code bit value, The coefficient difference may be set to a quantization threshold. For example, when the water code code bit value is 0, the coefficient difference of the watermark embedding region may be set to the first quantization threshold. When the watermark code bit value is 1, the coefficient difference of the watermark embedding region is equal to 0. It may be set to a value greater than the second quantization threshold or the second quantization threshold.

Next, in order to improve the quality of the watermark embedding image, the coefficient difference ( _fe ) aligned from the blocks in which the first watermark embedding region LH4 coefficient and the second watermark embedding region HL4 coefficient values are arranged in the order of the largest watermark is calculated. (S230). If the difference value between the first watermark embedding area (LH4) coefficient and the second watermark embedding area (HL4) coefficient is small, it can be used for coding with 0 bits, and the first watermark embedding area (LH4) coefficient and the second watermark insertion region (HL4) difference coefficients (1 _{έ, ¾)} value can be used to greater by 1 bit coding.

Then, two quantization thresholds (quant i tat ion thresholds) are set using the watermark code (20) data value () and the aligned coefficient difference (^!). The data value above the watermark code 20 is represented by a binary string value of zero or one.

5 is a diagram illustrating an example of a watermark image. First, as shown in FIG. 5A, an initial watermark image may be formed by using a predetermined image or code set by a scarlet person. However, in order to improve security, the initial watermark image data may be shuffled as shown in FIG. 5 (b) by using a random function, and the position of the watermark bit may be changed accordingly.

In this case, after the location and data of the initial watermark image data are mixed, an association key including location information of the watermark image data may be generated. this The association key may be used to restore the original watermark during the watermark extraction process.

The first quantization threshold () is obtained as shown in Equation 2.

[Equation 2]

0 means the number of data for which the watermark code data value (1 ^) is zero. The second quantization threshold () is obtained as in Equation 3.

[Equation 3]

A is a component for controlling the watermark embedding strength, and ^ denotes the number of data whose watermark code data value (:) is one.

As the value of S increases, the second quantization threshold also increases. Next, the channel to insert the watermark is selected using the sorted coefficient difference and the quantization threshold (S250). In this embodiment, the channel (^) having the smallest difference between the quantization threshold and the coefficient difference 4 _fc ) aligned as shown in Equation 4 is selected as the color channel to insert the watermark without inserting a watermark into all RGB channels. do. This inserts a watermark only into the selected channel (), resulting in more robustness. It can increase.

[Equation 4]

FIG. 6 is a flowchart illustrating an embodiment of the step of embedding the watermark code in the plurality of watermark embedding regions of the selected channel of FIG. 3. First, it is determined whether the data value w ^ is 1 or 0 of the watermark code 20 (S310). If the watermark code data value is 0, the first watermark embedding area coefficient and the second watermark embedding area are determined. Compare the coefficients (S320). According to the comparison result, as shown in Equation 5, the first _quantization is performed on the first watermark embedding region coefficient (low U watermark embedding region coefficient when C _i¾¾ J is greater than or equal to the second watermark embedding region coefficient ()). Threshold: Add the difference between () and the aligned coefficient difference ( _¾ ). On the contrary, if the second watermark embedding region coefficient (<: _∞ ,) is larger than the first watermark embedding region coefficient, the coefficient difference c) aligned with the first quantization threshold () in the second watermark embedding region coefficient 계수) Add the difference of (). [Equation 5]

< Denotes the difference between the first quantization threshold and the aligned coefficient difference. If the data value 1 of the watermark code 20 is 1, first, the sorted coefficient difference is compared with the second quantization threshold (S330). If the aligned coefficient difference information is greater than or equal to the second quantization threshold (), the existing first watermark embedding region coefficient value and the second watermark embedding region value are maintained as shown in Equation 6 (S332). .

[Equation 6]

If the aligned coefficient difference _fe is smaller than the second quantization threshold ₂ , the first watermark embedding region coefficient and the second watermark embedding region coefficient are compared (S340).

A second quantization threshold for the comparison according to claim 1 If the watermark embedding area coefficients second watermark embedding area coefficient (greater than or _½¾ J are the same, the first watermark embedding area coefficient _(¾fc J as shown in Equation (7) The difference between the coefficient coefficient () and the aligned coefficient difference () may be added, and the difference between the second quantization threshold value () and the aligned coefficient difference information ( _¾ ) may be subtracted from the second watermark embedding region coefficient (C _H½¾ ,). (S342). On the contrary, when the second watermark embedding region coefficient is larger than the first watermark embedding region coefficient,), the difference between the coefficient difference ( _& ) aligned with the second quantization threshold and the first watermark embedding region coefficient ( _¾ . Vf) may be subtracted, and a difference () between the second quantization threshold value ₂ and the aligned coefficient difference ¾ may be added to the second watermark embedding region coefficients C _H½ , _fc , (S344). [Equation 7], '< ^δ 2

〜 = C _Hh + vj

The watermark extraction apparatus 200 according to an embodiment of the present invention is used to extract the watermark code 20 from the watermark embedded image 30. The watermark embedded image 30 may be modified in transmission by potential attacks such as digital signal processing.

Referring to FIG. 7, first, the watermark extraction apparatus 200 performs discrete wavelet transform on the watermark embedded image to obtain wavelet coefficients from the watermark embedded image 30 (S410). The first watermark embedding area LH and the second watermark embedding area HL coefficients are grouped into blocks and blocked (S420). The block including the watermark information may be identified using an association key including the position information of the watermark image data after mixing the position and data of the initial watermark image data generated when the watermark is inserted. Since the extraction process cannot identify the block containing the watermark information unless the relevant key is used, the attacker cannot distinguish the block to which the watermark is applied and the block to which the watermark is not applied. ^'Is the following, differences of the first watermark embedding area coefficient and the second watermark embedding area coefficient calculated as shown in Equation 8 (S430).

[Equation 8]

The range of original difference coefficients is compressed and adjusted before calculating the reference value. Therefore, the original difference coefficient may be calculated as in Equation 9.

[Equation 9]

Where Τ is the average value of the coefficient difference between the watermark embedding regions, and may be calculated as shown in Equation 10.

[Equation 10]

After the N, the heroic reference value is calculated as a value that minimizes the variance of the original difference coefficient based on the Otsu method as shown in Equation 11 (S440).

[Formula 11]

Where *) denotes the variance of the original difference coefficient. And watermark by comparing the coefficient difference and the red-eye reference value as shown in Equation 12 Code 20 data values can be extracted.

[Equation 12] w. = ί ¹ ^ ^{≥ δ}

1 Ιθ otherwise

Finally, the original binary watermark image can be obtained by restoring the extracted watermark data value using the associated key.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements of those skilled in the art using the basic concepts of the present invention defined in the following claims are also provided. It belongs to the scope of rights.

Claims

[Claim]

[Claim 1]

A watermark embedding apparatus embeds a watermark in an input image _. Wavelet converting the input image for each color channel; extracting a plurality of watermark embedding regions from the wavelet transformed input image;

Selecting a color channel into which the watermark is to be inserted using bit values of the watermark image and coefficient differences between the plurality of watermark embedding regions; and

And inserting a watermark code composed of binary numbers into the plurality of watermark embedding regions extracted from the selected color channel.

[Claim 2]

In paragraph 1,

The plurality of watermark embedding regions

Selecting a channel for embedding the watermark, comprising a first watermark embedding region, and a second watermark embedding region;

Calculating a difference between the first watermark embedding region coefficient and the second watermark embedding region coefficient;

Determining a quantization threshold using the watermark insertion factor of the primary zone ^■, the watermark code data value, and

And selecting a channel having the smallest difference between the coefficient difference of the watermark embedding region and the quantization threshold as a watermark embedding channel.

[Claim 3]

In paragraph 2,

Computing a difference between the first watermark embedding region coefficient and the second watermark embedding region coefficient

Align the first watermark embedding region coefficients and the second watermark embedding region coefficients in ascending order and then calculate the sorted coefficient difference. Ow ᄇ ow.

[Claim 4]

In paragraph 3 '

The setting of the quantization threshold is

Setting a first quantization threshold, and setting a second quantization threshold having a value greater than the first quantization threshold;

ᄇ ᄇ o t

[Claim 5]

In claim 4,

Selecting the watermark embedding channel

The difference between the first quantization threshold and the aligned coefficient difference in a region where the data of the watermark code is 0, and

Selecting a channel having the smallest difference between the second quantization threshold and the aligned coefficient difference in an area of which data of the watermark code is 1 as a color channel for embedding a watermark.

[Claim 6]

In paragraph 1,

Extracting the plurality of watermark embedding regions

And extracting an LH region and an HL region into a watermark embedding region among the subbands obtained by wavelet transforming the input image for each color channel.

[Claim 7]

In paragraph 1,

The method of wavelet converting the input image for each color channel may include determining a singular number in consideration of a length and an insertion intensity of the watermark code and a degree of image quality degradation due to the insertion of the watermark code.

[Claim 8]

A device for embedding a watermark in the input image,

A wavelet transform unit for wavelet transforming the input image for each color channel, and a plurality of watermark embedding regions in the wavelet transformed input image. Watermark embedding area extraction unit,

A channel selector for selecting a color channel into which the watermark is to be inserted using bit values of the watermark image and coefficient differences between the plurality of watermark embedding regions; and

And a watermark embedding unit for embedding the watermark in the plurality of watermark embedding regions of the selected channel.

[Claim 9]

From the 18th paragraph,

The plurality of watermark embedding regions

And a first watermark embedding region and a second watermark embedding region, wherein the channel selector

Quantization for calculating a difference between the first watermark embedding region coefficient and the second watermark embedding region coefficient, and determining a quantization threshold using the coefficient difference of the watermark embedding region and the watermark code data value. A watermark embedding apparatus comprising a threshold value determining unit.

[Claim 10]

In claim 9,

The channel selector

And a channel having the smallest difference between the coefficient difference of the watermark embedding region and the quantization threshold as a watermark embedding channel.

[Claim 11]

In claim 9,

The channel selector

And sorting the first watermark embedding region coefficient and the second watermark embedding region coefficient in ascending order and then calculating the sorted coefficient difference.

[Claim 12]

In claim 9,

The quantization threshold is

A first quantization threshold, and having a value greater than the first quantization threshold A watermark embedding apparatus comprising a low 12 quantization threshold.

[Claim 13]

In claim 12,

The channel selector

And selecting a channel having the smallest difference between the second quantization threshold and the aligned coefficient difference in a region where the data of the watermark code is 1 as a color channel for embedding a watermark.

[Claim 14]

In claim 8,

The watermark embedding area extracting unit

And a watermark embedding region for extracting the LH region and the HL region from the subbands obtained by wavelet transforming the input image for each color channel.

[Claim 15]

In claim 8,

The wavelet transform unit

And determining the number of stages in consideration of the length and insertion strength of the watermark code and the degree of image quality deterioration due to the insertion of the watermark code.

[Claim 16]

A watermark extraction apparatus extracts a watermark from a watermark embedded image.

Obtaining a plurality of watermark embedding area wavelet coefficients by performing discrete wavelet transform on the watermark embedding image;

Blocking a plurality of watermark embedding region coefficients

Identifying a block including the watermark code using an associative key;

Calculating a plurality of watermark embedding region coefficient differences;

A value for minimizing dispersion of coefficient differences between the plurality of watermark embedding regions. Calculating a heroic reference value, and

And extracting the watermark code by using the reference value and the plurality of watermark embedding region coefficient differences.

[Claim 17]

In claim 16

And calculating original difference coefficients using average values of the plurality of watermark embedding region coefficient differences.

[Claim 18]

In claim 17,

[Claim 19]

In claim 18,

By inserting the reference and orient the plurality of watermark areas compares the coefficient ^difference. Extracting the watermark code value as 1 when the coefficient difference between the plurality of watermark embedding regions is greater than or equal to the adaptation reference value,

And extracting the watermark code value as 0 when the coefficient difference between the plurality of watermark embedding regions is smaller than the red-eye reference value.