KR20060041542A - Method of embedding and detecting watermark using dither modulation - Google Patents

Abstract

A semi-fragile watermarking method for preventing forgery and alteration of still images using dither modulation and discrete cosine transform is disclosed.

The watermark information can be inserted only in the coefficient at a specific position of the discrete cosine region, and the coefficient of any discrete cosine region can be modified in the spatial region, thereby improving the watermark embedding speed. Therefore, even a small resource such as a digital camera, a digital camcorder, a mobile phone, and the like, a semi-soft watermark embedding method for preventing forgery and alteration can be used.

In addition, dither modulation can be used to control the insertion strength of semi-weak watermarks and to increase the robustness against compression, making it possible to reliably detect forgery and forgery even when the compression ratio of the image is high.

Dither Modulation, Discrete Cosine Transform, Semi-Soft Watermark

Description

Watermark embedding and detection using dither modulation {METHOD OF EMBEDDING AND DETECTING WATERMARK USING DITHER MODULATION}

1 is a flow diagram illustrating a process of embedding a semi-soft watermark in an image in accordance with the present invention.

2 is a flowchart illustrating a process of generating a Discrete Cosine Transform (DCT) mask according to the present invention.

FIG. 3 is a diagram illustrating an example of a DCT mask generated through the process of generating the DCT mask of FIG. 2. FIG.

4 is a flowchart illustrating a process of detecting a semi-soft watermark from an image with a watermark embedded therein according to the present invention.

5A and 5B illustrate a method of determining whether an image is forged or forged using probability according to the present invention.

6 is a diagram illustrating an example of discriminating forgery from a forgery image.

7 is a view showing the degree of robustness to the compression of the watermark when using the watermark embedding and detection method according to the present invention.

※ Explanation of code about main part of drawing ※

100: first pseudonoise sequence (first watermark)

102: second pseudonoise sequence (second watermark)

The present invention relates to a method for embedding and detecting watermarks, and more particularly, to a method for embedding and detecting semi-fragile watermarks using dither modulation and Discrete Cosine Transform (DCT).

In recent years, with the rapid increase of Internet users, the trade of products through electronic commerce has been active, and this active e-commerce is an important procedure that should not be without digital certificate certification during the trade of products or online / offline transactions. Is rising.

In addition, the importance of images taken using devices such as digital cameras and digital camcorders is increasing. However, these digital contents are easy to forge and alter using various editing tools after being stored in a storage medium due to their digital properties. There is this. Accordingly, in order to protect the copyright and ownership of the authors who produce digital content and to prevent illegal copying of digital content, a technology that can distinguish whether or not digital content is forged is required. In order to prevent this, a fragile watermarking technique has been developed that can insert a watermark into digital content to detect forgery and forgery or to accurately detect a forgery and forgery.

In this case, the soft watermark refers to a technique of inserting a watermark into digital content to determine whether the data is forged or modulated, or to accurately detect the position of the forged or modulated data in the image.

On the other hand, since the size of the image data is generally large, it is common to compress and store these images. However, in the soft watermarking technology, even compressing an image is considered to be a forgery and alteration of the image. Therefore, the weak watermarking technique cannot detect forgery or forgery of the compressed image. Therefore, a semi-fragile watermarking technique is needed to solve this problem. In other words, it is a watermarking technique that is fragile for general external attacks but robust for compression.

In addition, the conventional watermark embedding method has a problem that it is difficult to insert a watermark quickly because many calculations are required because two quantization processes (that is, a compression process) are required for embedding and extracting the watermark.

It is an object of the present invention to provide a method for embedding a semi-soft watermark that is robust to compression in an image using dither modulation.

Another object of the present invention is to provide a method of embedding a semi-soft watermark in a portable device having a small resource such as a digital camera, a digital camcorder, a mobile phone, and the like by enabling fast watermark insertion using the linearity of discrete cosine transform. It is.                         

Still another object of the present invention is to prevent forgery and forgery of an image by detecting a watermark and determining whether the image is forgery or forgery through a stochastic method.

The present invention to solve the above problems, (a) dividing the image into blocks of a predetermined size; (b) obtaining a Discrete Cosine Transform (DCT) coefficient of the divided image; (c) performing dither modulation on the obtained DCT coefficients; And (d) multiplying the difference between the DCT coefficients before performing the dither modulation and the DCT coefficients after the dither modulation by a DCT mask and summing them with the divided image.

On the other hand, step (b), the DCT coefficient of the position determined according to the first pseudo-noise sequence is obtained, and step (c), it is preferable to perform dither modulation according to the second pseudo-noise sequence.

Further, the DCT mask is an inverse DCT coefficient for the location where the watermark is inserted, the watermark is a semi-fragile watermark that is weak against general external attacks but robust to compression, and the predetermined size is 8 The size of x8 and the length of the first pseudonoise sequence is preferably 12.

In addition, the present invention to solve the above problems, (e) dividing the watermark embedded image into blocks of a predetermined size; (f) performing a discrete cosine transform by identifying a position of a second pseudonoise sequence using the first pseudonoise sequence; And (g) determining whether the image is forged or forged after extracting the watermark.

On the other hand, step (g) compares the extracted watermark with the second pseudo-noise sequence, obtains a probability value for determining that the image has been modulated and a probability value for determining that the image has not been modulated, and calculates probability values obtained in the respective blocks. In addition, it is desirable to determine whether the image is forged or forged.

In addition, the predetermined size is 8x8, and the watermark is preferably a semi-soft watermark that is weak against general external attack but strong for compression.

Hereinafter, a method for embedding and detecting a watermark according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a flowchart illustrating a process of embedding a semi-soft watermark in an image according to the present invention.

First, the original image is divided into blocks of 8x8 size (S100). Next, a position for performing Discrete Cosine Transform (DCT) on the image divided into 8 × 8 blocks according to the first pseudonoise sequence 100 (first watermark) is determined (S102). In this case, if there is no first pseudonoise sequence, the position to perform the DCT transform is fixed.

Since the size of the block of the image segmented in step S100 is 8 × 8, the length of the first pseudonoise sequence corresponding to one DCT coefficient of the position at which the discrete cosine transform is to be performed is “6”. That is, the position is determined using information of 3 bits (0, 1, ..., 7) representing the horizontal position and information of 3 bits (0, 1, ..., 7) representing the vertical position. For example, if the input first pseudo noise sequence is "100110", the position of (4, 6) is selected among the 8 x 8 image blocks. In the present invention, it is preferable to select one coefficient at a low frequency and another coefficient at an intermediate frequency so as to be strong in compression, that is, since two watermarks are inserted in one block, the position at which the watermark is inserted is selected. To determine, a first pseudonoise sequence of length '12' per block is needed.

Next, the discrete cosine transform (DCT) is performed to obtain a DCT coefficient of the position determined in step S102 (S104). In this case, the equations of 8 × 8 two-dimensional discrete cosine transform (DCT) and inverse discrete cosine transform (IDCT) are

Where f (x, y) represents the input image, F (u, v) represents the result (DCT coefficient) of the discrete cosine transform, and coefficient C is

to be.

In the present invention, rather than obtaining the DCT coefficients at all positions of the image block of 8x8 size in step S104, only the DCT coefficients of the positions determined in step S102 are obtained, so that the amount of calculation is greater than obtaining the DCT coefficients of the entire 8x8 size. It is greatly reduced, thereby improving the watermark embedding speed. For example, as described above, if two DCT coefficients for the position determined in step S102 are obtained, the amount of calculation is reduced to 1/32 compared with the conventional technique of obtaining the DCT coefficients of the entire image block.

Next, after determining the watermark embedding intensity Δ in Equations 3 and 4 below (S106), the second pseudonoise sequence 102 (the second watermark) is applied to the DCT coefficients obtained in step S104. Accordingly, dither modulation is performed (S108).

In this case, the dither modulation formula is

는 디더 변조 후의 DCT 계수를 의미한다. 즉, 제 2 의사잡음 시퀀스 d_i 가 '0' 이면 수학식 3 에 따라 디더 변조하고, '1' 이면 수학식 4 에 따라 디더 변조한다.Where u, v (u = 0,1, .., 7; v = 0,1, .., 7) is the position of the DCT coefficient to which the watermark is to be inserted, and C _POS (u, v) is The DCT coefficient at position (u, v), Q means quantization. Δ denotes the watermark embedding strength, sign (C _POS (u, v)) denotes the sign of the DCT coefficient C _POS (u, v), and if the DCT coefficient is '0', sign (0) = 1 Set it. And d _i is a second pseudonoise sequence (second watermark),

Denotes a DCT coefficient after dither modulation. That is, when the second pseudo noise sequence d _i is '0', dither modulation is performed according to Equation 3, and when the second pseudonoise sequence d _i is '1', dither modulation is performed according to Equation 4.

Next, a watermark to be inserted is generated by multiplying the difference between the DCT coefficient before dither modulation and the DCT coefficient after dither modulation obtained through dither modulation in step S108 by a DCT mask obtained in advance in the spatial domain, and dividing it in step S100. The image is summed with the 8 × 8 image (S110). In this case, the 'DCT mask' means the inverse DCT coefficient for the position at which the watermark is inserted. In step S110, the present invention provides the linearity of the discrete cosine transform, that is,

By using, it is possible to change the portion of the DCT coefficients to be changed in the spatial domain without performing the DCT transformation and the inverse DCT (IDCT) transformation.

Hereinafter, a process of generating a DCT mask will be described with reference to the flowchart of FIG. 2.

First, the image is divided into 8 x 8 blocks (S200). Next, if the DCT coefficient at the position (n, m) of the DCT region is to be changed, only the coefficient at the position (n, m) is set to '1' and the rest is set to '0' (S202). Next, inverse discrete cosine transform (IDCT) is performed (S204) to generate a DCT mask (S206).

FIG. 3 is a diagram illustrating an example of a DCT mask generated through the process of generating the DCT mask of FIG. 2. For example, as shown in (a) of FIG. 3, when the DCT coefficient at position (2,7) is to be changed in a block of size 8 × 8, only the coefficient at position (2,7) is set to '1'. And set the rest to '0'. Inverse discrete cosine transform is then performed using Equation 1 above to generate a DCT mask as shown in FIG. 3 (b), and when the DCT mask is added to image data in the spatial domain and then DCT transformed, By the linearity of the above-described DCT transform, the data values shown in FIG. 3A are added to the DCT transformed data of the original image, resulting in the increase of only one data at the position (2,7) by one.

Therefore, in the present invention, only the DCT coefficient to which the watermark is to be inserted can be changed, and fast spatial domain processing can be performed using the DCT mask generated through the preliminary work as shown in FIGS. 2 and 3.

That is, in step S110 of FIG. 1 using the DCT mask, the watermark obtained by adding the inverse DCT result by the magnitude of the difference between the DCT coefficient before dither modulation and the DCT coefficient after dither modulation to the divided image (8x8). You can get the inserted image.

to the next. Steps S102 to S110 are repeated for all 8x8 blocks (S112).

Hereinafter, a process of detecting a watermark according to the present invention will be described.

4 is a flowchart illustrating a process of detecting a semi-soft watermark from an image having a watermark embedded therein according to the present invention.

First, the image into which the watermark is inserted is divided into 8 x 8 blocks (S400).

Next, for each block, the position of the second pseudonoise sequence is checked using the first pseudonoise sequence of step S102 to perform DCT transformation to obtain a DCT coefficient (S402). Next, a watermark is extracted from the DCT coefficient obtained in step S402 and then compared with the inserted second pseudo noise sequence to calculate a probability value that is determined to be modulated and a probability value that is not modulated (S404). In this case, the probability value is

If pn _i = 0

,

,

if pn _i = 1

는 제 2 의사잡음 시퀀스가 '0' 일 경우 DCT 계수값을 디더 변조해서 구한 값이며,

은 제 2 의사잡음 시퀀스가 '1' 일 경우에 디더 변조해서 구한 값이다. 또한

는 위조 변조 여부를 판단하고자 하는 위치 (m, n) 의 DCT 계수를 의미한다.Where pn _i denotes a second pseudonoise sequence (second watermark), L denotes a low frequency region, M denotes an intermediate frequency region, and i denotes a block of 8 × 8 size. In addition, P _Ti represents a probability of being true, that is, a probability value that is determined to be unmodulated, and P _Fi represents a probability of being 'false', that is, a probability value determined to be modulated.

Is a value obtained by dither-modulating a DCT coefficient value when the second pseudo noise sequence is '0',

Is a value obtained by dither modulation when the second pseudonoise sequence is '1'. Also

Denotes the DCT coefficient of the position (m, n) to determine whether the forgery modulation.

Next, with reference to Figs. 5A and 5B, a method of determining whether or not an image is forged or false based on the probability value calculated in step S404 will be described.

5A shows an image 500 divided into an image 502 of 8 × 8 size blocks and one DCT coefficient 504 of the 8 × 8 DCT coefficients.

For example, as shown in FIG. 5B, when the DCT coefficient value 504 of one of the blocks 502 of FIG. 5A is a, the second pseudonoise sequence is quantized to A when '0', and Assuming that 1 'is quantized to B, if the second pseudonoise sequence is' 0', α is a probability of 'true', that is, an image is not modulated, and β is a probability of 'false', that is, modulation It is likely that

Next, the sum of probability values corresponding to the n blocks is determined to determine whether to forgery or modulate (S406). Since image information may be different in the process of compressing the image data, in order to determine whether the forgery and forgery are more accurate, the coefficients of the n blocks are summed to determine whether the forgery or forgery is stochastic. By comparing the probability values of true and false thus obtained, as shown in Equation 7 below,

If the true probability value is greater than the false probability value, it is determined that there is no forgery or alteration of the image, and vice versa, the image is forgery or alteration. In this case, n means the number of blocks used to determine whether or not forgery among the blocks divided in step S400.

Hereinafter, referring to FIG. 6, an example in which a semi-soft watermark is inserted into an original image, and a part of the image is forged and forged and then forged or forged is determined.

For example, after modulating the original image of FIG.

1: Replace with original image 2: Delete contents

3: graffiti 4: changing colors

5: Delete part of the picture 6: Copy the image of another part

7: Rotate image 8: Detach and paste candles from other images

9: detach and paste a butterfly from another image

10: Write a letter

By using the forgery and modulation detection method described above with reference to FIGS. 4 and 5, the modulated portion is displayed to be distinguished from the original image, as shown in FIG. 6, so that the forgery and position of the image can be accurately detected. It can be seen that.

Hereinafter, when the watermark embedding and detection method is used in an image using dither modulation according to the present invention, the degree of robustness of the watermark with respect to compression will be described with reference to FIG. 7. FIG. 7 shows a semi-soft watermark detection error when JPEG compression is performed after inserting a semi-soft watermark with an insertion intensity (Δ) of 20 for 15 images of a size of 1280 × 1024. . In this case, the average PSNR was 50.45 dB, and as shown in FIG. 7, when the semi-soft watermark embedding method using dither modulation is used, the robustness of the compression is large, so that even if the compression ratio of the image is increased, it is possible to determine whether the forgery and modulation are reliable. It can be seen that.

According to the present invention, it is possible to insert a semi-soft watermark, which is robust against compression, by using dither modulation, so that even if the compression ratio of the image is increased, it is possible to determine whether forgery and forgery are reliable.

In addition, the present invention can insert watermark information only in the coefficients of the discrete cosine region of a specific position, and the coefficient of any discrete cosine region in the spatial domain can be modified by using the linearity of the discrete cosine transform, so that the rapid watermark insertion is possible. Accordingly, even a small resource such as a digital camera, a digital camcorder, a mobile phone, and the like, a semi-soft watermark embedding method can be used.

In addition, the present invention can detect the presence or absence of the forgery and forgery of the image by using a probabilistic method after detecting the watermark in the image in which the semi-soft watermark is inserted.

In the above, the preferred embodiments of the present invention have been illustrated and described, but the present invention should not be limited to the specific preferred embodiments described above, and the technology to which the present invention belongs without departing from the gist of the present invention of the appended claims. Various modifications can be made by those skilled in the art, and such modifications are within the scope of the claims.

Claims (12)

In the method of embedding a watermark in an image, (a) dividing the image into blocks of a predetermined size; (b) obtaining a Discrete Cosine Transform (DCT) coefficient of the divided image; (c) performing dither modulation on the obtained DCT coefficients; And and (d) multiplying the difference between the DCT coefficients before performing the dither modulation and the DCT coefficients after the dither modulation by a DCT mask, and adding them to the divided image. The method of claim 1, In the step (b), the DCT coefficient of the position determined according to the first pseudo noise sequence is obtained. The method of claim 1, In the step (c), the dither modulation is performed according to the second pseudo noise sequence. The method of claim 1, And the DCT mask is an inverse DCT coefficient for the position at which the watermark is inserted. The method of claim 1, And the watermark is a semi-fragile watermark that is weak against general external attacks but strong against compression. The method of claim 1, And the predetermined size is 8 × 8. The method of claim 2, The predetermined size is the size of 8 × 8, And the length of the first pseudo-noise sequence is 12. A method for detecting a watermark in an image into which a watermark has been inserted by the watermark embedding method according to any one of claims 1 to 7, (e) dividing the watermarked image into blocks of a predetermined size; (f) performing a discrete cosine transform by identifying a position of a second pseudonoise sequence using the first pseudonoise sequence; And and (g) extracting the watermark with respect to the discrete cosine transform coefficient obtained in step (f) and determining whether the image is forged or forged. The method of claim 8, In the step (g), the extracted watermark is compared with the second pseudo-noise sequence to obtain a probability value for determining that the image has been modulated and a probability value for determining that the image has not been modulated, thereby determining whether the image is forged or forged. Watermark detection method. The method of claim 9, In the step (g), it is determined whether the image is forged or forged by summing probabilities obtained from each of the divided blocks. The method of claim 8, The predetermined size is a watermark detection method, characterized in that the size of 8 × 8 The method of claim 8, Wherein said watermark is a semi-weak watermark that is weak against general external attacks but robust to compression.

Images