JP2020088469A - Digital watermarking apparatus and method - Google Patents

Abstract

To obtain a digital watermarking apparatus and method that are difficult to remove a watermark and resistant to the removal against attack from a third party by an invisible digital watermarking method.SOLUTION: In an invisible and highly secure digital watermarking method that is robust against attack from a third party, image data is divided into blocks, and an embedding pattern is selected from a different dot pattern p0, q0 exhibiting a green noise characteristic for each block, and its inversion pattern by a selection list determined in advance by a random number sequence to perform embed watermark with the selected pattern.SELECTED DRAWING: Figure 4

Description

The present invention is a digital watermark method for embedding information in digital image data for the purpose of copyright protection and prevention of forgery, and in particular, an invisible digital watermark apparatus and method having high resistance to watermark removal by an attack from a third party. It is about.

In the current digital information society, it has become possible for many people to share information by duplicating information, and the society has greatly developed. However, due to its convenience, illegal copying and distribution of other people's copyrighted works has led to cases such as copyright infringement. With regard to images, due to the high image quality of digital cameras and printers in recent years, it is possible to easily obtain a copy that is exactly the same as the original image. The result is that it promotes a malicious criminal act called an act.

In such a situation, invisible digital watermark technology that embeds other information in the image information, for example, author information to protect the copyright and warns against illegal copying Has developed. For example, for the purpose of copyright protection, this digital watermark technology not only alerts users by embedding the copyright owner, URL, contact information, terms of use, etc. in the copyrighted work of documents and images. If it is illegally used, it can be traced. Further, in the case of illegal copying, at the same time as giving a warning, it is possible to stop copying in a copying apparatus or output a blackened image. As described above, digital watermarking has begun to be widely used as a security measure such as copyright management/protection and monitoring of illegal copying.

The digital watermark as a security measure must be robust and must not be easily erased or removed. This is because if the watermark information can be removed or erased by a simple operation, it can be escaped from the eyes of tracking or monitoring by illegal use.

The digital watermark as digital data does not deteriorate or disappear as long as it is copied as it is. However, when the image is edited or processed by conversion such as enlargement, reduction, or rotation, or image compression such as JPEG is performed, the embedded information is easily lost. In addition, when cropping a part of an image and using it, the embedded watermark information is broken. In order to avoid the above situations and protect copyrights and prevent forgery, it is necessary to have a robust digital watermark technology that can withstand these editing and processing.

Furthermore, it is assumed that a third party removes the watermark information for illegal use (third party attack). By analyzing and estimating the embedding algorithm with various methods and various images, the watermark information is removed from the embedded image and returned to the original image, thereby enabling unauthorized use. Therefore, the digital watermark algorithm needs to protect against such an attack.

Therefore, the digital watermark method needs to have robustness (robustness) to image editing/processing and attack from a third party. On the other hand, in order to make it strong, it is necessary to increase the embedding strength (gain), but in general, a large gain causes deterioration of image quality, and an artifact is visually recognized. Image quality is important for commercial purposes such as stock photo distribution on the Internet, and invisible digital watermarks that do not cause image quality deterioration are effective as a measure to protect copyrights. Is requested.

Usually, the tolerance and the image quality are in a trade-off relationship. That is, if the resistance is increased, the image quality is deteriorated, and if the deterioration of the image quality is reduced, the resistance is decreased. Digital watermarks for copyright protection require algorithms and techniques that achieve both image quality and durability.

As a digital watermark method satisfying such a demand, there is a method of embedding watermark information in a frequency space. This is a method of performing frequency conversion on an image and embedding embedded information in a specific frequency band. Since the embedded information is distributed over the entire image in real space, it is visually inconspicuous.

As a digital watermark method in a frequency space, Non-Patent Document 1 proposes a method in which an image is DCT-transformed (discrete cosine transform) and watermark information is embedded in an intermediate frequency band in the DCT space. The embedded information is dispersed and diluted in the entire image in the real space. When the gain of the embedding strength is increased, the artifact is strongly expressed, and the shape of the artifact changes depending on the embedding position. When the embedding position is embedded at a plurality of points of a radial intermediate frequency from the position of the DC component (0 frequency), diagonal lines are taken in the real space, and long lines of low frequency noise are visually recognized and the image quality is deteriorated.

On the other hand, in the digital watermark method for embedding after Wavelet conversion, an image is subjected to DWT (Discrete Wavelet Transform) conversion, and a watermark is embedded in the Wavelet component. When the gain is increased, the pattern embedded in the inversely transformed image appears as an artifact. The reason for this is that the Wavelet transform is a spectral intensity transform that preserves the spatial position of the image, so that it appears at the embedding position without being dispersed over the entire image in real space. The one embedded in the LL component appears stronger than the one embedded in the HH component. Also, in multi-layer conversion, the more embedded it in the upper layer, the stronger the artifact. For this reason, methods such as embedding and dispersing in the edge portion are taken, but the image dependency is strong. If the gain is reduced, the image quality will not be a problem, but if high tolerance is required, it will be an important problem.

Therefore, the method of embedding the green noise pattern shown in Patent Document 1 in image data (hereinafter referred to as the green noise diffusion method) has been previously proposed as a digital watermark method with high resistance and high image quality. The spectrum of the green noise pattern is confined in a specific band f_(max) to f_min, and its main frequency can be set to a frequency band that is difficult to recognize by human visual characteristics. For this reason, when used for a digital watermark, when the embedded image is observed at a distance of clear vision, the dot pattern has low responsiveness and is difficult to be recognized. Therefore, even if the embedding strength (gain) is increased and the editing/processing is tougher, no artifact is felt, and high durability can be maintained while maintaining high image quality.

In the green noise diffusion method, an embedding operation is performed by dividing an image into blocks and superimposing a green noise dot pattern corresponding to embedding bits on the image data. The watermark is extracted by performing FFT (Fast Fourier Transform) on the embedded image in block units and extracting the embedded information from the spectrum distribution. Since the embedding is a real space and the extraction is a frequency space, which is a so-called asymmetric type, and the green noise pattern is a random dot pattern, it is difficult to erase or tamper with the watermark itself.

However, it is assumed that the watermark will be removed by a so-called collusion attack that specifies the watermark pattern from the original image, the pattern of the flat portion, the plurality of images, and the like. By taking the difference between the embedded image and the original image, the embedded pattern can be extracted. Even if there is no original image, the watermark pattern is partially extracted in block units from the image embedded in the flat part of the image (for example, sky landscape or sea landscape), and the entire embedded pattern is estimated from multiple blocks. It is possible to To avoid this problem and increase security, it is necessary to further increase resistance to these attacks.

JP 2018-182471

T. Mizumoto, K. Matsui; “Study on printing uptake resistance of digital watermark using DCT”, The Institute of Electronics, Information and Communication Engineers A, Vol J85-A, No.4, pp.451-459 (2002).

The challenge is to obtain a digital watermark method that is resistant to image editing and processing, and that has little deterioration in image quality even if it is strongly embedded, and that has strong resistance that the watermark information is not removed even when attacked by a third party. is there. Another issue is that the author and purchaser can remove the watermark as needed. Another challenge is to make the system secure so that damage is not spread over a wide area against the loss of keys.

In order to solve the above problems, the present invention divides an image into blocks, embeds a digital cluster dot pattern showing a green noise characteristic in each block by superimposing the random cluster dot pattern on the original image.

The image data is divided into blocks, and in each block, the bit information (0 or 1) of the watermark to be embedded is selected from two basic dot patterns p0 and q0 that exhibit green noise characteristics and have different spectral shapes, and their inversion patterns. And select the embedding pattern from the selection list, multiply this pattern by gain and superimpose it on the image data to generate a watermark embedding image,
The watermark information is extracted by dividing the image data in which the watermark is embedded into blocks, and extracting the embedded bit information from the shape of the Fourier spectrum pattern of the blocks.

Here, the inverted dot pattern is generated from the two basic dot patterns p0, q0,
p1,q1: upside down dot pattern,
p2,q2: Horizontally inverted dot pattern,
p3,q3: Vertical and horizontal inverted dot pattern,
p4,q4: p0,q0 negative dot pattern,
p5,q5: p1,q1 negative dot pattern,
p6,q6: p2,q2 negative dot pattern,
p7,q7: p3,q3 negative dot pattern,
Which is all or a part of each of 8 dot patterns pi, qi (i=0,1,...,7) including 7 inversion patterns
The selection list shows the selection order of pi and qi dot patterns.

The basic dot patterns p0 and q0 are composed of k patterns each having the same spectral shape but different dot patterns. All of the 8×k dot patterns including the inverted dot pattern, or one of them. Part as a dot pattern for embedding,
The selection list shows the selection order of those dot patterns.

The selection list is a random number generator that randomly selects a plurality of dot patterns to be used.

Further, such a digital watermark can remove the embedded watermark with a key composed of i basic dot patterns pi and qi, gain, embedded information and a selection list.

If the basic dot pattern qi is generated by using the basic dot pattern pi, the digital watermark is embedded with a key composed of i basic dot patterns pi and gain, embedded information and a selection list. It is possible to remove the watermark.

According to the present invention, it is possible to embed a watermark that is robust against an attack from a third party. Since there are a plurality of watermark embedding dot patterns including the reverse dot pattern, it is difficult to estimate the basic dot pattern from the difference from the original image or the flat portion of the image.

In addition, when a plurality of basic dot patterns are used, the dot patterns are further increased by including the reverse pattern. Moreover, since the selection list is random, there is no regularity. Therefore, it is more difficult to estimate the dot pattern.

On the other hand, the key for removing and removing usually requires all dot patterns, but the reverse dot pattern is generated from the basic dot pattern, so the key information is basic dot patterns p0, q0, gain, and embedded information. And only the selection list is needed, and the key for removing is small.

Furthermore, when the basic dot pattern q0 is generated from p0, the key information is only p0, gain, embedded information, and the selection list, and the key for removal has a smaller capacity.

With such a key, it is possible to remove the watermark from the image in which the watermark is embedded and restore the original image (original image). By changing the key for each distribution image, it is useless for other images. It is also possible to embed new secondary copyright information as a digital watermark using the key.

FIG. 1 is a diagram of an apparatus for embedding and extracting a digital watermark of the present invention, Diagram showing the delivery of watermarked images on the Internet, In the figure of basic dot pattern for embedding, (a) 32×32 has the characteristics of a vertically elongated elliptical ring, (b) is a 90° rotated version of it, and (c) is a 64×64 vertically elongated elliptical ring. Has the spectral characteristics of, and (d) is the one rotated 90 degrees. The figure showing a basic dot pattern and a reverse dot pattern. An image with different gains embedded and an enlarged view of part of it. (a) gain=0.0625, (b) 0.125, (c) 0.25. The figure showing the processing flow of watermark embedding. The figure showing the gain automatic adjustment processing flow at the time of watermark embedding. The figure showing the processing flow of watermark extraction. The figure of the mask pattern of watermark extraction. The figure which shows the mask process of watermark extraction. The figure which shows the reliability improvement by the mask process of watermark extraction. In the figure of watermark embedding, extraction, and removal, (a) is an image with a watermark embedded (gain=0.1875), (b) is a spectral image for watermark extraction, and (c) is an image with the watermark removed. The figure showing the different dot pattern of the same spectrum. The figure showing two basic dot patterns and a reverse dot pattern.

FIG. 1 is a block diagram of a digital watermark system for embedding and extracting information according to the present invention. In the computer 3 of the copyright holder, image data having a copyright is stored in a data memory 7 such as a hard disk. The image data is subjected to image processing by the image processing program in the program memory 6 using the CPU 11, ROM 4, RAM 5, etc., and is displayed on the monitor 8. A scanner 1 and a printer 2 are connected to the computer 3, and the processed image can be output from the printer and the image can be read from the scanner 1. Such image processing increases a calculation load when handling a large size image, and thus may include a processing board for performing parallel processing or increasing the speed of a GPU or the like.

In order to embed confidential information or copyright information in an image, character information such as the author's name, date, and URL is entered from the keyboard 9. The number of characters that can be embedded differs depending on the image size and block size. When the block size is 512 pixels×512 pixels and the block size is 64 pixels×64 pixels, a maximum of 8 bytes is possible, and when the block size is 32 pixels×32 pixels, a maximum of 32 bytes is possible. As far as ASCII code is concerned, these are 8 characters and 32 characters, respectively.

FIG. 2 shows an Internet distribution processing procedure in one operation mode. The image in which the watermark information is embedded is distributed via the Internet 12 from the computer of the distributor or the copyright holder 16 (13). The purchase applicant 17 who sees it informs the distributor of the purchase request (14). After the predetermined procedure, the distributor distributes the private key for embedding and removing the watermark (15). Based on the contract with the distributor, the purchaser can remove the digital watermark embedded in the sent image and edit the image.

FIG. 2 shows an Internet distribution processing procedure in one operation mode. The image in which the watermark information is embedded is distributed via the Internet 12 from the computer of the distributor or the copyright holder 16 (13). The purchase applicant 17 who sees it informs the distributor of the purchase request (14). After the predetermined procedure, the distributor distributes the private key for embedding and removing the watermark (15). Based on the contract with the distributor, the purchaser can remove the digital watermark embedded in the sent image and edit the image.

A third party who sees the image can confirm that the image is copyrighted by the watermark extraction software. The watermark extraction software has no confidentiality and can be freely downloaded from the homepages of distributors and copyright holders and widely published, and can be used by anyone. This allows a third party to know the owner of the image, copyright information, contact information, etc., which also serves as a warning/prevention of unauthorized use.

On the other hand, the copyright holder and the secondary copyright holder can trace and monitor the image data as the copyrighted work by the watermark. If a third party uses it illegally, the copyright holder or secondary copyright holder will read the watermark information from the image using the monitoring software, and it will be his own work, and the third party will use it without permission. If so, it is possible to detect.

The key has a one-to-one correspondence for each image to be embedded. Therefore, the copyright holder uses a different key for each image so as not to be misused. This key cannot be used to remove watermarks on other images. Even if a key is lost or a key is acquired by a collusion attack by a third party, the damage is only to the one image in question, and the watermark cannot be removed from other images. At this time, the watermark reading software can be commonly used even if the keys are different. Also. Since the keys can be issued almost infinitely, the probability that they will be the same key is extremely low.

Hereinafter, detailed description will be given along with examples.
Here, the dot profile creation algorithm showing the green noise characteristic used in the present invention will be described first.
Now, assuming the dither matrix size to be obtained is R×R (R=2^m), the blackening rate representing the gradation is g (0≦g≦1:g=1 is all black, g=0 is all white) To do. The dot profile at the blackening rate g and point (x,y) is p(x,y), and the dot profile for intermediate density (g=1/2) is obtained as follows.
(STEP1) First, (R^2)/2 random dots (initial state is white noise) are generated by a pseudo-random number generator and set as p(x,y). At this time, the dot profile in the initial state can be changed by changing the SEED value of the pseudo random number generator. Two-dimensional Fourier transform of such dot profile is performed to obtain P(fx,fy).
(STEP2) P(fx,fy) is multiplied by the filter D(fx,fy) to obtain a new P'(fx,fy). Here, D(fx,fy) is a filter having a value in the region where the radial frequency fr is fmin to fmax.
(STEP3) Inverse Fourier transform is performed on P'(fx,fy) to obtain a multi-valued point profile p'(x,y).
(STEP4) The error function e(x,y)=p'(x,y)-p(x,y) is obtained, and white and black are inverted in the order of large error at each pixel position.
(STEP5) The above operation is repeated until the error is within the allowable amount, and finally the dot profile p0 of g=1/2 is obtained.

Here, the setting of the radial frequencies fmax and fmin will be described. The frequency due to the average dot spacing at g=1/2 at fmax and fmin of filter D(u,v) is
fo=√g・fn=√(1/2)・fn
And fmax and fmin are based on fo
a≡(fmax fo)/fn
b ≡ (fmin fo)/fn
The parameter (a,b) is defined as Here, fn is the Nyquist frequency. Dot patterns with different cluster sizes can be obtained by changing (a, b).

As an example, the dot profile when R=32, (a,b)=(-1/16,-5/16), and ellipticity of 1.5 is obtained. For watermark embedding, two types of patterns corresponding to embedding bits (0,1) are required. Therefore, p0 is a vertically long ellipse and q0 is a horizontally long ellipse pattern. The dot pattern p0 and the spectral characteristics are shown in FIG. 3(a). The ellipticity of 1.5 means that the filter D(fx,fy) is enlarged 1.3 times in the y-axis direction. The imaginary part is 0 because the spectral distribution is symmetrical with respect to the x and y axes. .. Figure (b) shows the dot pattern q0. This is generated individually, but in order to reduce the memory capacity of the key to be described later, p0 rotated by 90° is substituted. The spectrum is also rotated by 90°. The same figures (c) and (d) show p0 and q0 when R=64. The spectral shape becomes clearer.

Next, a watermark embedding algorithm for image data will be described. To embed the watermark information, the color image data is converted into Y, Cb, Cr and the watermark information is embedded in the luminance Y. It is ideal to embed in Blue (yellow at the time of printing) because it is the most visually unrecognizable and ideal, but it is difficult to accurately separate the printed image or monitor image from the image read by the camera or scanner. ..

Subsequently, the image data is divided into R pixel×R pixel blocks. This should be the same size as the dot pattern size. Normally, 64 pixels×64 pixels are used as the block size, but if a large amount of information is to be embedded, setting the block size to 32 pixels×32 pixels enables embedding of information four times as large.

The watermark information is embedded in blocks in the real space. Now, if the image data is i(x,y), the embedding strength is gain, and the watermarked image is w(x,y),
When embedded bit = 0: w(x,y)=i(x,y)+gain・p0(x,y) (1)
When embedded bit = 1: w(x,y)=i(x,y)+gain・q0(x,y) (2)
Becomes. As for the watermark information, a character string is converted into a bit string in advance, and the watermark information is sequentially embedded in block units. Here, p0 and q0 are binary values of (1,0), but are set to (1/2,-1/2) in order to preserve the average luminance.

Since p0 and q0 are random dots, the boundary between them is not noticeable. The dot pattern exhibiting green noise characteristics is also used as an FM screen in printing, and because it is a dispersive dot and has excellent uniformity, it is visually uniform and does not feel graininess.

When embedding only p0 and q0 in the image data, there is a risk that the embedding pattern will be reproduced by a third party analyzing the pattern. Since the p0 and q0 patterns are repeatedly embedded, the original pattern can be estimated by observing them in detail. Especially in an image with a uniform background such as the sky or the sea, the dot pattern is easy to identify, and the estimated pattern can be obtained based on the information obtained from each block.

For this reason, the present invention uses a large number of embedding patterns so that they cannot be easily estimated. FIG. 4 is a diagram in which an inverted dot pattern is generated from one green noise dot pattern p0, q0, so that the dot pattern is propagated into eight types of patterns.
When P0,q0 is the basic dot pattern, the following inverted dot pattern is generated from p0,q0 as follows.
p1,q1: upside down dot pattern,
p2,q2: Horizontally inverted dot pattern,
p3,q3: Vertical and horizontal inverted dot pattern,
p4,q4: negative/positive inversion dot pattern of p0,q0,
p5,q5: Negative/positive inversion dot pattern of p1,q1,
p6,q6: negative/positive inversion dot pattern of p2,q2,
p7,q7: p3,q3 negative/positive inversion dot pattern,
FIG. 4 shows the inverted dot pattern, pi, qi (i=0, 1,..., 7) in the case of 32×32. The watermark information is embedded in the image using all or part of this dot pattern.

For embedding watermark information, the order of embedded dot patterns is determined by the selection list L. The selection list L is, for example, a list for selecting which dot pattern number to use, and is determined by a random number before embedding. Now, assuming that all pi,qi (i=0,1,2,...,7) are used, a random number from 0 to 7 is generated by the random number generator by the number of blocks. For example, obtain the following selection list L.
L={0,7,5,2,4,7,1,5,6,1,6,4,2,7,6,0,3,1,6,0,3,2,7,5 ,…} (3)
The selection pattern list is used to determine the dot pattern number i used in the block order. That is, if the block number is n,
i=L(n)
And pi,qi are determined. For example, if i=0, p0,q0, and if i=7, p7,q7. Using this pi, qi, the embedding is for each block,
When embedded bit = 0: w(x,y)=i(x,y)+gain pi(x,y) (1')
When embedded bit = 1: w(x,y)=i(x,y)+gain qi(x,y) (2')
Becomes

Since the embedding is performed in block units, the image size can be any size. The amount of data (maximum number of bits) N that can be embedded in an image is the image size for W pixel × H pixel image data,
N=int(W/R)×int(H/R) (4)
Given in. Here, int() represents rounding down to the nearest whole number. The number of embeddable characters is int(N/8) in case of ASCII characters. Usually, the target image for which copyright protection is desired is an image of full HD size (1920 pixels × 1080 pixels) or more, and in a full HD image, N = 480 bits = 60 bytes, which is sufficient for embedding copyright information. .. However, when a small size image requires a sufficient amount of information, if R=32 is embedded, for example, in the case of 512 pixels×512 pixels, an information amount of N=256 bits=32 bytes can be embedded. is there.

The tolerance of the embedded image can be increased by increasing the gain. However, the dot pattern becomes visible when the gain is increased. FIG. 5 shows an image when the gain is changed and embedded, and a partially enlarged image thereof. The figure (a) is embedded with gain=0.0625, and the embedded information is almost unknown visually. However, watermark extraction is error prone. On the other hand, in FIG. 6C, the watermark is embedded with gain=0.25, and the extraction of the watermark is almost 100%, but the dot structure becomes noticeable. As a result of the experiment, it was found that when the gain is 0.125 to 0.25, the green noise dot pattern is not conspicuous due to the low-pass effect due to the visual characteristics and is resistant and practical.

FIG. 6 shows a watermark embedding processing flow. It is assumed that the dot pattern to be embedded already exists. First, a watermark information bit string wm(i,j) for embedding and a selection list L(n) are prepared (20). Here, i represents the i-th character to be embedded and j represents the j-th bit, but it is converted into a one-dimensional bit string in advance. Subsequently, the image data is divided into R×R blocks (21), and dot patterns are synthesized in block units. First, start with the first block n=1 (22). Next, the selected dot pattern number i is obtained from the selection list L(n) from the block number n, and the embedded dot patterns pi and qi are determined (23). For embedding, it is judged whether the embedding bit is 0 or 1 (24), if it is 1, embedding qi (25), and if it is 0, embedding pi (26). Subsequently, it is judged whether or not the block is completed (27), and if not completed, the block number is incremented by 1 (28) and the process returns to (23). The process ends when all blocks have been embedded.

The embedded watermark information may not be extracted because the embedded gain is small. This occurs when the spatial frequency of the corresponding block of the original image is high. FIG. 7 shows a flow for automatically adjusting gain. First, watermark information wm(i,j) is embedded in all blocks of an image with gain=g (30). After all blocks are embedded, the watermark is read next (32). Here, the extracted watermark information wm'(i,j) and the watermark information wm(i,j) before embedding are compared to determine whether they match. At the same time, it is judged whether the reliability wmrel'(i,j) of the extracted bit, which will be described later, is greater than or equal to a certain threshold value (33). If the watermark information does not match or if the reliability is equal to or less than the threshold value even if they match, a constant value Δg is added to gain (34) and the watermark is embedded again. Since the reliability can be quantified, it is possible to automatically embed with the optimum watermark strength as described above. This is to allow a margin because the watermark information may not be read if the reliability is less than or equal to the threshold value even if the watermark information matches.

Next, the extraction of the watermark will be described.
The watermark information is extracted by obtaining the spectral image in block units after correcting the size and inclination of the read image. Fourier transform of equation (2),
F{w(x,y)}=F{i(x,y)} + gain・F{pi(x,y) or qi (x,y)} (5)
Becomes However, F{ }represents the Fourier transform. Usually, F{i(x,y)} is the spectrum of the original image, localized near 0 frequency, and F{pi(x,y) or qi(x,y)} is the ring-shaped spectrum surrounding it. Becomes Therefore, there is little overlap between the two and it is easy to separate them.

FIG. 8 shows a processing flow of watermark extraction. First, the input image is divided into R×R blocks (41), and FFT is performed on the first block (42). Subsequently, histogram equalization is applied to the obtained spectral image (43). This is to improve the contrast of the spectral pattern and at the same time standardize the quantification of the reliability, which will be described later. Subsequently, a masking process described later is performed (44) to obtain watermark information wm'(i,j) and reliability wmrel'(i,j). It is judged whether or not all the blocks are finished (46). If not finished, the process goes to the next block and the same process is performed again.

FIG. 9 shows a mask pattern as an identifying device that extracts watermark information by pattern recognition from the spectrum distribution after Fourier transform. A vertically and horizontally long elliptical ring-shaped mask that corresponds to the spectral characteristics of a dot pattern for embedding. In the figure, the black area has a value of 1, and the white area has a value of 0. This mask is overlaid on the watermarked spectrum, and it is judged whether the bit is 0 or 1 from the difference of the integrated luminance value of the overlapping portion. That is, in FIG. 10, outputs Q0 and Q1 of the following integrated luminance values are obtained for the spectral pattern W(x,y) of the block (40).
Q0= (1/Z) ΣM0(x,y)・W(x,y)
Q1=(1/Z)ΣM1(x,y)・W(x,y) (6)
From such output,
When Q0>Q1, extraction bit＝wm'(x,y)=0
When Q1>Q0, extraction bit＝wm'(x,y)=1
(41). At the same time, the reliability is obtained (42) as follows.
wmrel'(i,j)=|Q0-Q1|(7)
The larger the value of the reliability, the higher the reliability of the extracted bit.

The improvement of the accuracy and reliability of watermark information will be described with reference to FIG. The maximum number of embedded bits N is given by equation (4) depending on the size of the image. Now, assuming that the number of bits to be embedded is n and n<N, the bit strings can be embedded redundantly. In the extraction of the watermark information in FIG. 8, up to N consecutive watermark information wm' and reliability wmrel' were obtained regardless of the number of bits (without considering the repetition of the character string). A highly reliable one is selected from the continuous character strings to obtain watermark information up to n.

First, when the sth character number (0<s≦N) and t of the embedded character string are bit numbers (0≦t≦7), the remainder k modulo the number of characters n is k=smod n When (50) and the (s,t)th reliability wmrel' is wmrel'(s,t)>wmrel(k,t), the reliability is high.
wm(k,t)=wm'(s,t)
wmrel(k,t)=wmrel'(s,t) (8)
(52), the watermark information is replaced. This is performed for all bits and all blocks until s reaches N, that is, the number of embedded bits N ends.

By extracting the embedded block with low reliability in this way by subdivision, extraction with high reliability becomes possible. The blocks divided into four all have the same spectral characteristics, so if even one high reliability is extracted, the entire block is upgraded to that reliability. In this way, watermark extraction with high reliability can be performed on the printed image.

Next, the watermark removal will be described.
The watermark information can be removed by obtaining the original image by knowing the embedding pattern. That is, from equations (1') and (2'),
When embedded bit = 0: i(x,y)= w(x,y)-gain pi(x,y) (9)
When embedded bit = 1: i(x,y)= w(x,y)-gain qi(x,y) (10)
Then, by knowing the embedding patterns pi(x,y), qi(x,y) and gain, the embedding information and the selection list, it is possible to restore the current image. This information can be done by receiving a "secret key". In order to completely restore the original image, it is necessary to limit the dynamic range in advance in order to avoid the overflow and underflow of the embedded image.

FIG. 12 is an image in which eight characters “20181118” are embedded in an image of 256 pixels×256 pixels using the dot pattern of FIG. (A) shows a diagram when embedding with gain=0.1875, (b) shows a spectrum for extraction, and the embedded watermark information could be accurately extracted. (C) is an image in which the watermark is removed from the embedded image of (a) with a key, and the original image is reproduced.

The dot pattern included in such a secret key may be only the basic dot patterns p0 and q0, and the inverted dot pattern can be generated from this basic dot pattern. Therefore, the key size is 4096 bits (512 bytes) for both p0 and q0 when R=64, and the total capacity is the sum of the embedded information and the selection list, but the embedded information and selection list are proportional to the number of characters. However, the dot pattern capacity is approximately the key capacity, which is approximately several tens to 100 bytes.

If the basic dot pattern of q0 is used by rotating the basic dot pattern of p0 by 90°, the key is the basic dot pattern p0, gain, embedded information and selection list, and the key capacity is further reduced. ..

Next, the attack resistance of this method will be described. Attacks on digital watermarks include, for example, signal enhancement (sharpness adjustment), noise addition, filtering (linear, non-linear), lossy compression (JPEG, MPEG), transformation (rotation, scaling). These attacks are attacks on luminance information, and embedding by a set of dispersed dots as in this method can be maintained unless attacks in the resolution direction (for example, a strong low-pass filter), and are generally robust. I can say. Tolerance can be further increased by increasing the gain.

Next, the possibility that a third party may illegally detect and remove the embedded pattern will be described. In order to specify the embedding pattern, it is generally assumed that the watermark pattern is specified from the original image in which the watermark is not embedded or a plurality of embedding images. If only basic dot patterns are used, the same pattern is repeated many times because they are used repeatedly, so the dot pattern of the original block size can be estimated by connecting the partial patterns obtained from each block. ..

In order to avoid this, it is possible to have many basic dot patterns. This is because it becomes difficult to specify the pattern by using a plurality of dot patterns. However, the capacity of the key increases and it becomes difficult to handle. Therefore, the number of patterns can be increased and the key size can be prevented from increasing by generating the inverted pattern from the basic dot pattern.

For the loss or theft of the key, it is not possible to remove the watermark of another image with the same key by using different basic dot patterns for each image. FIG. 13 shows different dot patterns with the same spectral characteristics. These are obtained by changing the seed value of the initial random number when the green noise pattern is generated. The spectral distribution is the same, the extraction software is common, but the dot profile is different. When R=64, only (_4096^)C_2048 different patterns are possible, and the probability of having the same pattern is extremely low, so the safety is high.
The copyright holder provides the purchaser with a dot pattern obtained from different random numbers as a secret key. Since this key is different for each purchaser, the watermark cannot be removed from the images of other purchasers.
As described above, the present invention provides a highly resistant digital watermark that eliminates attacks from third parties.

Next, a method of increasing the number of basic dot patterns to obtain a digital watermark with stronger attack resistance is shown. In FIG. 14, two basic dot patterns (pattern 1 and pattern 2) are used as basic dot patterns.
As shown below, eight patterns p0 to p7 including seven inverted patterns from pattern 1 are obtained. Eight patterns p8 to p15 are similarly obtained from the pattern 2.
p0, p8: Basic dot pattern
p1, p9: Upside down dot pattern,
p2, p10: Horizontally inverted dot pattern,
p3, p11: Vertical and horizontal inverted dot patterns,
p4, p12: Negative/positive inversion dot pattern of p0, p8 patterns,
p5, p13: negative/positive inversion dot pattern of p1, p9,
p6, p14: p2, p10 negative/positive inversion dot pattern,
p7, p15: p3, p11 negative/positive inversion dot pattern,
Similarly, an inverted dot pattern is generated from q0, q8 having a spectrum obtained by rotating p0, p8 by 90°.

When all patterns are used, the selection list L selects which of the 16 patterns from p0 to p15 and q0 to q15 to use, so a random number generator blocks all numbers from 0 to 15 Only generate for a few minutes. For example, obtain the following selection list L.
L={4,10,5,12,9,7,14,15,6,11,6,4,2,7,6,8,13,1,6,0,3,12,7,… } (3)
The selection pattern list is used to determine the dot pattern number i used in the block order. That is, if the block number is n,
i=L(n)
And pi,qi are determined. For example, if i=4, p4,q4, and if i=10, p10,q10. Using this pi, qi, the embedding is for each block,
When embedded bit = 0: w(x,y)=i(x,y)+gain pi(x,y) (1')
When embedded bit = 1: w(x,y)=i(x,y)+gain qi(x,y) (2')
Becomes

In the above, the case where the basic dot patterns p0 and q0 are two each has been described, but it is assumed that the basic dot patterns p0 and q0 are each made up of k patterns. Since a maximum of seven inversion patterns are generated for each basic dot pattern, there is a maximum of 8×k in total of the basic dot pattern and the inversion dot pattern, and all or a part of them are used as the embedding dot pattern.
The selection list shows the selection order of those dot patterns.

As in the case of the first embodiment, the watermark is extracted by first dividing the image in which the watermark is embedded into R×R blocks, performing FFT for each block, and performing mask processing to obtain watermark information and reliability. Even if the number of dot patterns is increased and embedded, the extraction software does not need to be changed.

Increasing the number of dot patterns is effective in avoiding attacks from third parties. However, the size of the key increases and it becomes difficult to handle. In the present invention, in order to reduce the capacity of the key, an inversion pattern is generated from the basic dot pattern and only the basic dot pattern information is used as the key information. Therefore, the compact key facilitates handling such as distribution and storage.

Furthermore, the key size is further reduced by generating the basic dot pattern q0 having a 90° rotated spectrum shape from the dot pattern of p0. However, reducing the number of basic dot patterns makes it easier to attack, so it is necessary to select an optimal method according to the level of resistance.

The digital watermarking method of the present invention has been described above. However, since the method embeds a dot pattern exhibiting a green noise characteristic, it is possible to embed a dot pattern with little visual discomfort and strongly edit the image. Strongly resistant to attacks from people. Further, since the original image can be restored by the key information, it is effective not only for copyright protection but also for various applications by linking the printed image and electronic data.

Further, since the embedding of the main watermark is performed in blocks in the real space, there is no limitation on the image size such as an image of a large size, and the watermark can be processed as fast as the dither method. For this reason, it is possible to embed moving images in real time, and it can be expected to be applied to surveillance cameras and drive recorders.

This watermark embedding and extraction method has no problem even if the algorithm is disclosed. All confidentiality is aggregated into a dot pattern, which is stored as a private key. Even if the key is lost or stolen, if a different dot pattern is used for each image, there is one key for one image in principle, so the watermark is used for other images using that key. It cannot be removed. For this reason, the security is extremely high, and disclosure of algorithms, standardization, and disclosure of extraction software do not pose a problem.

1 is scanner, 2 is printer, 3 is computer system, 4 is ROM, 5 is RAM, 6 is program memory, 7 is data memory, 8 is monitor, 9 is keyboard, 10 is communication function, 11 is CPU, 12 is Reference numeral 13 is the Internet, 13 is distribution of image data, 14 is notification of purchase request, 15 is transmission of secret key, 16 is computer of copyright holder, and 17 is computer of purchaser.

Claims (6)

In the digital watermarking method of embedding watermark information in digital image data, the image data is divided into blocks, and in each block, two basic dot patterns p0, q0 and their inversion patterns showing green noise characteristics and different spectrum shapes Select the embedding pattern from the bit information (0 or 1) of the watermark to embed and a selection list, multiply this pattern by gain and superimpose it on the image data to generate a watermark embedding image,
The watermark information extraction is performed by dividing image data in which a watermark is embedded into blocks and extracting embedded bit information from the shape of a Fourier spectrum pattern of the block. The inverted dot pattern is generated from the two basic dot patterns p0, q0,
p1,q1: upside down dot pattern,
p2,q2: Horizontally inverted dot pattern,
p3,q3: Vertical and horizontal inverted dot pattern,
p4,q4: negative/positive inversion dot pattern of p0,q0,
p5,q5: Negative/positive inversion dot pattern of p1,q1,
p6,q6: negative/positive inversion dot pattern of p2,q2,
p7,q7: p3,q3 negative/positive inversion dot pattern,
Which is all or a part of each of 8 dot patterns pi, qi (i=0,1,...,7) including 7 inversion patterns
The digital watermarking apparatus and method according to claim 1, wherein the selection list indicates a selection order of pi and qi dot patterns. The basic dot patterns p0 and q0 are k patterns each having the same spectral shape but different dot patterns, and all or a part of the 8×k dot patterns including the inverted dot pattern. As a dot pattern for embedding,
The electronic watermarking apparatus and method according to claim 1, wherein the selection list indicates a selection order of those dot patterns. 4. The digital watermarking apparatus and method according to claim 3, wherein the selection list is a selection order of a plurality of dot patterns used, which is randomly generated by a random number generator. 5. The digital watermark according to claim 4, wherein the embedded watermark can be removed by a key composed of i basic dot patterns pi and qi, gain, embedded information and a selection list. Digital watermarking apparatus and method. In the digital watermark, the basic dot pattern qi is generated from the basic dot pattern pi, and the key is composed of i basic dot patterns pi, gain, embedded information, and a selection list. The electronic watermarking apparatus and method according to claim 4.