KR20110091986A - A method of inserting and extracting digital watermark in the wavelet area by using wavelet transformation and boundaries data - Google Patents

Abstract

PURPOSE: A water marking method and water mark extraction method in a wavelet area using wavelet conversion and border information are provided to determine an area to insert a watermark by reflecting all shape detection in a energy distribution and a space area in a frequency area. CONSTITUTION: A digital image obtains a wavelet image by converting a multiple wavelet conversion(S10). A first candidate area is selected after calculating energy per unit area by dividing one among high frequency sub band of the wavelet image as a first unit area(S20). A second candidate area is selected after calculating alerting information per unit area by dividing the digital image as a second unit area(S30). Final candidate area is extracted by logically multiplying the first candidate area and the second candidate area(S40). By selecting one among the low frequency sub ban of the wavelet image, a watermark is inserted to an area of a sub band selected by corresponding to the final candidate area(S50). The wavelet image which is substituted as a sub band in which the watermark is inserted is restored(S60).

Description

Watermarking method and watermark extraction method using wavelet transform and boundary information {A method of inserting and extracting digital watermark in the wavelet area by using wavelet transformation and boundaries data}

The present invention relates to a watermarking method and a watermark extraction method in a wavelet region in which a watermark is inserted after a watermark is inserted using a tree structure between subbands configured in the wavelet transform region and image shape information in a spatial region. It is about.

In particular, the present invention divides a subband having a high frequency into a plurality of regions in a wavelet region, obtains a candidate region for watermarking by using an average value of energy and total energy for each block, and boundaries of an image obtained in a spatial region. The present invention relates to a watermarking method and a watermark extraction method in a wavelet area in which a candidate area into which a watermark is to be inserted using information is obtained, and a final watermarking selection area is obtained using these two candidate areas.

In addition, the present invention relates to a watermarking method and a watermark extraction method in a wavelet region in which a watermark is randomly substituted using LFSR and random random numbers and then inserted into a watermarking selection region.

With the rapid development of the Internet, the movement and storage of digital data via wired / wireless is becoming easier and faster. However, digital data is easy to copy and alter, and the original and copy are difficult to distinguish. With the spread of illegal copying and distribution of digital data, the issue of intellectual property has long been highlighted as an important social problem. In order to solve this problem, digital watermarking is a technology that can provide a basis for protecting intellectual property rights and claiming copyright. Currently, many researches and practical applications have been made.

  In the early 1990s, research was started with the introduction of watermarking for digital images. Since 1995, full-scale research on digital watermarking has been conducted. Watermarking, which was mainly studied in the space domain at the beginning, is shifting the research domain by applying it in the frequency domain. Although watermarking is more resistant to attack than the method applied in the space domain, the watermark insertion position cannot be accurately selected due to the frequency characteristics. There was a downside.

However, the introduction of the wavelet region, which has both the frequency characteristics and the spatial characteristics, makes watermark insertion more efficient. Watermarking in the frequency domain inserts a watermark by changing the frequency coefficients. The method of extracting a significant coefficient in the frequency domain using a discrete cosine transform to insert a watermark, and a watermark in phase using a discrete Fourier transform. Insertion methods and methods using discrete wavelet transform have been mainly studied. Recently, many methods using correlations of subbands using zerotree after wavelet transform have been developed.

However, inserting a watermark by energy information extracted from the tree structure may be robust to attack, but has a problem of invisibility.

An object of the present invention is to solve the problems as described above, watermarking method in the wavelet region to improve the invisibility while increasing the invisibility by introducing shape detection in the spatial domain as well as energy distribution in the frequency domain And a watermark extraction method.

In particular, the present invention first predicts the energy distribution of the low frequency region according to the energy distribution of the high frequency band in the subband tree structure generated after the wavelet transform and detects the shape information of the image in the spatial region, The present invention provides a watermarking method and a watermark extraction method in a wavelet region for selecting a location in a subband for inserting a watermark using information from.

In addition, an object of the present invention is a watermarking method and a watermark extraction method in a wavelet region in which a watermark image is randomly and irregularly inserted through a secret key and a linear feedback shift register (LFSR) in a region selected for watermark insertion. To provide.

In order to achieve the above object, the present invention relates to a watermarking method in a wavelet region in which a watermark is inserted into a wavelet region of a digital image using wavelet transformation and boundary information, and (a) multilevel wavelet transformation of the digital image. Obtaining a wavelet image; (b) selecting a first candidate region by dividing one of the high frequency subbands of the wavelet image into a first unit region to calculate energy for each unit region; dividing the digital image into a second unit area, calculating boundary information for each unit area, and selecting a second candidate area; (d) extracting a final candidate area by ANDing the first candidate area and the second candidate area; (e) selecting one of low frequency subbands of the wavelet image and inserting the watermark in a region of the selected subband corresponding to the final candidate region; And (f) restoring the wavelet image substituted with the subband in which the watermark is inserted.

In addition, the present invention provides a watermarking method in the wavelet region, when the D-level wavelet transform is performed, the size of the first unit region is 2 ^D-1 × 2 ^{D- 1} , and the size of the second unit region is 2 and ^D 2 × ^D, the first high-frequency sub-band to be used in selection of a candidate area is an LH sub-band _1, the low-frequency sub-band that is watermarked is characterized in that the LH _sub-D inverse.

In addition, in the watermarking method in the wavelet region, in step (c), boundary information is detected by using a Canny method using a Gaussian mask and a Sobel mask. It is characterized by.

In addition, the present invention provides a watermarking method in the wavelet region, wherein in step (b), if the energy of the first unit region is greater than the first threshold value, the bit value is set to 1 and the bit value is smaller than the first threshold value. Is set to 0, and the first candidate area is configured as a bit plane.In step (c), if the number of boundary information of the second unit area is greater than the second threshold value, the bit value is set to 1 and is greater than the second threshold value. If it is small, the bit value is set to 0, and the first candidate region is configured as a bit plane. In step (e), the watermark is inserted only in the portion of the low frequency subband corresponding to the portion where the bit value of the final candidate region is 1. Characterized in that.

In addition, in the watermarking method in the wavelet region, in step (e), the coordinate value x (i, j) of the low frequency subband is set to x '(i, j) by Equation 1. The watermark is inserted by being replaced.

[Equation 1]

Where "·" is a logical product, β is a scaling component, k (i, j) = 0 or 1,

W (i, j) is the watermark.

In the watermarking method of the wavelet region, in the step (f), the insertion position of the watermark is replaced by a LiDear feedback shift register (LFSR) method.

According to another aspect of the present invention, there is provided a watermarking method in a wavelet region, wherein the digital image is subjected to three-level wavelet transform.

The present invention also relates to a watermark extraction method in a wavelet region in which the watermark is extracted from an embedded image in which a watermark is inserted into an original image by the watermarking method in the wavelet region. Extracting a low frequency subband through wavelet transform; obtaining a difference between the extracted low frequency subband coefficients and the wavelet transform coefficients of the original image; And (i) extracting a watermark using the difference and the bit plane of the final candidate area.

As described above, according to the watermarking method and the watermark extraction method in the wavelet area according to the present invention, by defining the area to insert the watermark reflecting both the energy distribution in the frequency domain and the shape detection in the spatial domain As a result, it is possible to increase the invisibility while being robust against various attacks.

In particular, according to the watermarking method and the watermark extraction method in the wavelet area according to the present invention, having a robustness against a general image processing attack, it can be used for applications such as copyright protection or ownership verification in the field of image processing Obtained.

1 is a block diagram of an entire system for implementing a watermarking method and a watermark extraction method in a wavelet area according to the present invention.
2 is a flowchart illustrating a watermarking method in a wavelet area according to an embodiment of the present invention.
3 is a diagram illustrating a calculation process of a cumulative multiplier according to an embodiment of the present invention.
4 is a diagram illustrating a subband structure in which a three-level wavelet transform of a digital image is performed according to an embodiment of the present invention.
5 is a flowchart illustrating a method of selecting a first candidate region in a wavelet image according to the present invention.
6 is a flowchart illustrating a method of detecting edges in a digital image according to the present invention.
FIG. 7 is a diagram illustrating a configuration of replacing a position of a watermark using a linear feedback shift register (LFSR) according to an embodiment of the present invention.
8 is a flowchart illustrating a watermark extraction method in a wavelet area according to an embodiment of the present invention.
9 is a flowchart illustrating a method of extracting a watermark using a difference in coefficients and a bit plane of a final candidate area according to an embodiment of the present invention.
Explanation of symbols on the main parts of the drawings
11,12: computer terminal 20: network
30: watermarking system 40: watermark extraction system
70: digital video 80: embedded video

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the drawings.

In addition, in describing this invention, the same code | symbol is attached | subjected and the repeated description is abbreviate | omitted.

First, the configuration of the entire system for implementing the watermarking method and the watermark extraction method in the wavelet area according to the present invention will be described with reference to FIG.

As shown in FIG. 1, the entire system for implementing the present invention includes a watermarking system 30 generating an embedded image 80 by inserting a watermark in a digital image 70, and water in the embedded image 80. It consists of a watermark extraction system 40 which extracts a mark.

As an example, the watermarking system 30 or the watermark extraction system 40 may be implemented as a software tool installed and executed on the computer terminals 11 and 12. Computer terminals 11 and 12 are conventional computing terminals. Examples of computer terminals include a PC, a notebook computer, a PDA, a server, and the like, and the detailed description thereof will be omitted since the configuration and operation principle thereof are well-known techniques commonly used in the art.

As another embodiment, the watermarking system 30 or the watermark extraction system 40 may be made of a single control circuit and made of chips.

The watermarking system 30 inserts and distributes the watermark into the digital image 70. The inserted image 80 in which the watermark is inserted is distributed online or offline.

The watermark extraction system 40 extracts the watermark into the embedded image 80 and checks the copyright by comparing the extracted watermark.

Next, a watermarking method in the wavelet area according to an embodiment of the present invention will be described with reference to FIGS. 2 and 3.

As shown in Figures 2 and 3, the watermarking method according to the present invention comprises the steps of (a) obtaining a wavelet image by performing a multi-level wavelet transform of the digital image (S10); (b) selecting a first candidate region by dividing one of the high frequency subbands of the wavelet image into a first unit region to calculate energy for each unit region (S20); (c) dividing the digital image into a first unit area and calculating boundary information for each unit area to select a second candidate area (S30); (d) extracting a final candidate area by ANDing the first candidate area and the second candidate area (S40); (e) selecting one of the low frequency subbands of the wavelet image and inserting the watermark in an area of the selected subband corresponding to the final candidate area (S50); And (f) restoring the wavelet image substituted with the subband in which the watermark is inserted (S60).

Hereinafter, an example of performing three-level wavelet transform of a digital image will be described. The following description can also be applied to the D-level wavelet transform.

In general, the watermarking method consists of two parts. The first is to find the position to insert the watermark, and the second is to insert the watermark at the selected position. First, information for selecting a position where a watermark is to be inserted at the same time in a wavelet area and a space area is obtained. Based on the obtained information, a position to insert a watermark in the wavelet area is selected.

First, a wavelet transform is performed on a digital image by multilevel (or three level) wavelet transform (S10), and one of the high frequency subbands of the wavelet image is divided into a first unit region to calculate energy for each unit region. One candidate area is selected (S20).

A frequency component having important information of the LH ₃ subband in the wavelet region LH ₁ subband is predicted. This is based on the assumption that there is a high correlation between the two subbands. In the spatial domain, shape information constituting an image is obtained using boundary information in the image. The area for inserting watermarking is selected based on the information in these two areas. The watermark is inserted in the LH ₃ subband in the wavelet region.

The watermark embedding method according to the present invention starts from the fact that the correlation between the high frequency subband and the low frequency subband is very large. In general, techniques using the correlation between subbands are a method of searching for important coefficients from low frequency subbands (LH ₃ , HL ₃ , and HH ₃ ) to high frequency subbands (LH ₁ , HL ₁ , and HH ₁ ), while the present invention The watermarking method follows a method of predicting and raising from a high frequency subband to a low frequency subband. The structure of the subbands after the three-level wavelet transform (three-level wavelet transform) is shown in FIG. 4. The watermark is assumed to be inserted into the LH ₃ subband for two reasons. This is because LH ₃ is generally the most resistant to attack and highly invisible.

Next, a first candidate region is selected from the wavelet image (S20). The specific method will be described in more detail with reference to FIG. 5.

As shown in FIG. 5, first, the LH ₁ subband is divided into a first unit region having a size of 4 × 4, and each region is calculated as an average of squares, that is, an energy in a frequency domain (S21).

If the size of the entire image is N × M, the size of LH ₁ is (N / 2) × (M / 2). The average value of the energy values for the 4 × 4 size region is calculated and named T _seg , and the threshold value is set (S22). That is, this value is a value obtained, the average of the energies calculated for the unit area of 4 × 4 unit ₁ in the LH sub-band to sub-band LH _1.

After scaling the threshold value, the energy values of each of 4 × 4 areas are compared and selected as a watermarking candidate area in the wavelet area (S23). The watermarking candidate region at this time is called a first candidate region. The watermarking candidate region (or first candidate region) in the wavelet region has a size of (N / 8) × (M / 8) and has a bit plane shape. Comparing the energy values and configuring the bit plane is relatively simple. If the energy value is larger than the scaled threshold αT _seg , the bit value of the bit plane is set to 1 and the bit value is set to 0. Where α corresponds to the scaling component and has a value between 0 and 1.

Next, boundary detection is performed on the digital image 70 (S30). The specific method will be described with reference to FIG.

The key is generated according to the energy distribution in the frequency band by wavelet transform, which includes the characteristics of various noises. Therefore, the edge detection algorithm is introduced to prevent key generation by background noise and to use Mach band effect in space domain. In other words, when the watermark is inserted, the watermark is used to determine the watermarking candidate area using the HVS.

The present invention detects a boundary using a Canny method using a Gaussian mask and a Sobel mask. A method of selecting a watermarking target region (or second candidate region) in a spatial region is as follows.

As shown in FIG. 6, first, a noise is removed using a Gaussian mask (S31), and a boundary is found using a Sobel filter (S32). In operation S33, a binary image representing boundary information is generated using the previously obtained threshold value. Next, after dividing the binary image into 8 × 8 unit areas, the number of elements having a value of 1 in the unit area is obtained (S34). Then, how much boundary information is included in the unit area is obtained by comparing the number of 1s and the threshold number of _edges (T _edge ). If the number of 1s is greater than T _edge , it is set to bit 1. Otherwise, it is set to bit 0. The bits are collected to form a bit plane, which is referred to as a watermarking candidate region (or second candidate region) in the spatial region (S35). The watermarking candidate area in the spatial area also has the same size (N / 8) × (M / 8) as LH ₃ .

Next, the watermarking candidate area in the wavelet area and the watermarking candidate area in the space area having the same size of (N / 8) × (M / 8) are each composed of bit planes. The final watermarking selection region (or final candidate region) is obtained through the AND of these two candidate regions (S40).

Next, a watermark is inserted in the region of the low frequency subband corresponding to the final candidate region (S50).

As shown in FIG. 7, before the watermark is inserted, the position of the watermark is replaced using a linear feedback shift register (LFSR) as a method for protecting the watermark. The watermark is replaced by generating a random address consisting of the x- and y-axes using the parallel output of the LFSR.

Watermarking is performed using the generated bit plane of the watermarking candidate region and the encrypted watermark (S50), and the watermarking process is completed through an inverse wavelet transform process (S60). Watermarking is applied at LH ₃ subbands. The watermark is a gray image having a value of 0 to 255, and the size is preferably (N / 16) × (N / 16). However, the size is not limited thereto.

Insertion of watermarking may be defined as in Equation 1 below.

[Equation 1]

Where ".&Quot; represents a logical product and β is a scaling component representing the toughness of the watermark. k (i, j) represents a watermarking candidate area. If the value is 1, the watermark is inserted, and if it is 0, it is not inserted. W (i, j) is a watermark, and the value is added or subtracted according to the value. The image with all the watermarks inserted is reconstructed into the image of the spatial domain through the inverse lifting conversion and the watermarking process is completed.

Next, a watermark extraction method in the wavelet area according to an embodiment of the present invention will be described with reference to FIG. 8.

To extract the embedded watermark, the following process is performed. After the wavelet transform is performed using the lifting (S70), a difference between the coefficient of the LH ₃ subband and the wavelet transform coefficient of the original image is taken (S80). The watermark is extracted using the information on the watermarking candidate region and the information of the LFSR initial value (S90). The watermark extraction process is shown in FIG. W '(i, j) corresponds to the extracted watermark.

As mentioned above, although the invention made by this inventor was demonstrated concretely according to the said Example, this invention is not limited to the said Example and can be variously changed in the range which does not deviate from the summary.

The present invention is applicable to the development of a watermarking system for inserting a watermark after determining the insertion position of the watermark using a tree structure between subbands configured in the wavelet transform region and image shape information in the spatial region. In particular, the present invention may be used for applications such as copyright protection or ownership verification in the field of video that requires robustness against general image processing attacks.

Claims (8)

A watermarking method in a wavelet region in which a watermark is inserted into a wavelet region of a digital image using wavelet transform and boundary information,
(a) multi-level wavelet transforming the digital image to obtain a wavelet image;
(b) selecting a first candidate region by dividing one of the high frequency subbands of the wavelet image into a first unit region to calculate energy for each unit region;
dividing the digital image into a second unit area, calculating boundary information for each unit area, and selecting a second candidate area;
(d) extracting a final candidate area by ANDing the first candidate area and the second candidate area;
(e) selecting one of low frequency subbands of the wavelet image and inserting the watermark in a region of the selected subband corresponding to the final candidate region; And
and (f) restoring the wavelet image substituted with the subband into which the watermark is inserted.
The method of claim 1,
When the D-level wavelet transform is performed, the size of the first unit region is 2 ^D-1 × 2 ^{D- 1} , and the size of the second unit region is 2 ^D × 2 ^D , which is used to select the first candidate region. The high frequency subband is an LH ₁ subband, and the low frequency subband into which the watermark is inserted is a LH _D subband.
The method of claim 1,
In the step (c), the watermarking method in the wavelet region, characterized in that for detecting the boundary information by using a Canny method using a Gaussian mask and Sobel mask.
The method of claim 1,
In step (b), if the energy of the first unit region is greater than the first threshold value, the bit value is set to 1, and if it is less than the first threshold value, the bit value is set to 0, and the first candidate region is configured as a bit plane. ,
In the step (c), if the number of boundary information of the second unit area is greater than the second threshold value, the bit value is set to 1, and if it is less than the second threshold value, the bit value is set to 0, and the first candidate area is set to the bit plane. Consist of,
In the step (e), the watermarking method in the wavelet region, characterized in that the watermark is inserted only in the portion of the low frequency subband corresponding to the portion of the bit value of the last candidate region.
The method of claim 4, wherein
In the step (e), the watermark is inserted by substituting the coordinate value x (i, j) of the low frequency subband into x '(i, j) by [Equation 1]. Watermarking method.
[Equation 1]

Where "·" is a logical product, β is a scaling component, k (i, j) = 0 or 1,
W (i, j) is the watermark.
The method of claim 1,
In the step (f), the insertion position of the watermark is a watermarking method in the wavelet region, characterized in that the replacement by a LFSR (LiDear feedback shift register) method.
The method according to any one of claims 1 to 6,
And a three-level wavelet transform of the digital image.
A watermark extraction method in a wavelet region for extracting the watermark from an embedded image having a watermark embedded in an original image according to any one of claims 1 to 6,
(g) extracting a low frequency subband through the wavelet transform of the inserted image;
obtaining a difference between the extracted low frequency subband coefficients and the wavelet transform coefficients of the original image; And
(i) extracting a watermark using the difference and the bit plane of the final candidate area.

Images