JP4942633B2 - Digital watermarking system - Google Patents

Description

  The present invention relates to a digital watermark system for embedding watermark information in image content.

  Digital watermarking is a technology that embeds information so that it cannot be perceived by human eyes or ears using the spatial correlation or temporal correlation of image content. Depending on the nature of the content, information is detected from the embedded content. Since it is often impossible to restore as complete information, a method of reducing the false detection rate using an error correction code is used.

  FIG. 8 is a block diagram showing a configuration of a conventional digital watermark system. In FIG. 8, the conventional digital watermark system includes a digital watermark embedder 100 and a digital watermark detector 200.

  The digital watermark embedder 100 is provided with an error correction encoder 101, an embedding position determiner 102, an embedding pattern generator 103, and an adder 104.

  The digital watermark detector 200 is provided with a subtractor 201, a detection data generator 202, and an error correction decoder 203.

  Next, the operation of the conventional digital watermark system will be described with reference to the drawings.

  In the digital watermark embedder 100, error correction coding such as a repetitive code in which each information bit is repeatedly made into redundant data by the error correction coder 101 with respect to embedded information (watermark information) 1 such as a distributed person's ID. To embed data 101d. Here, for example, when the embedding information 1 is 16 bits and the embedding data 101d is 16000 bits, it is a code for repeating each bit 16000/16 = 1000 times.

  The embedding position 102d is determined by the embedding position determiner 102 using a random number or the like with the key 2. The embedding pattern generator 103 obtains an embedding pattern 103d by setting -1 when the redundant bit is 0 and +1 when it is 1. Then, an embedded image 11 is obtained by adding the embedded pattern 103 d to the original image (image) 10 by the adder 104.

  The post-embedding image 11 is generally distributed, but since the ID of the person to be distributed is included as watermark information and it is known to whom the image is distributed, the person who wants to copy illegally and distribute it secondarily deteriorates the image quality. It tries to erase watermark information by an attack with few. The post-embedding image 11 is attacked by the attacking device 12, and the post-attack image 13 is obtained and distributed illegally.

  The digital watermark detector 200 obtains the post-attack image 13 that has been illegally distributed, and obtains a detection pattern 201 d that is a difference obtained by subtracting the original image 10 from the post-attack image 13 by the subtractor 201. Here, data “1” is determined when the detection pattern 201d is +, and “0” is determined when the detection pattern 201d is −.

  According to the embedding position 102d generated at the time of embedding, the detection data generator 202 sequentially extracts the detection data 202d including error redundant bits, and the error correction decoder 203 performs error decoding to obtain detection information 3. In the case of the error correction code repeated 1000 times, “1” is exceeded when “1” exceeds half 501 and “0” is exceeded when “0” exceeds half 501. Error decoding. When there is no attack, the embedded information 1 is detected as it is, so that the detection information 3 is accurate information.

  However, here, there is a case where the detector has the original image 10 itself, but for the sake of simplicity, the original image 10 is possessed and the data of the original image 10 can be used.

  FIG. 9 is a diagram showing a distribution of change amounts of each pixel after embedding digital watermark data by the conventional digital watermark system. In FIG. 9, the horizontal axis represents the amount of change from the original image, and the vertical axis represents the distribution. The area where data “1” is embedded has a change of +1 with respect to the original picture, and the area where data “0” is embedded has a change of −1 with respect to the original picture, and these are distributed almost symmetrically. The change is 0 in the non-embedded area.

  The attack is performed on the original image 10 by, for example, randomly adding a numerical value. When it is detected that ± 2 is randomly added to all pixels in a weak attack, it is “1” if the data is positive, “0” if it is negative, but this follows the polarity when ± 2 is added. The embedded data cannot be detected at all, and normal information cannot be detected.

  For example, Patent Document 1, Patent Document 2, and Patent Document 3 are prior art documents for digital watermarking.

JP 2000-106624 A JP 2004-104494 A JP 2006-025409 A

  In the conventional digital watermarking system as described above, information is embedded by giving a change of a fixed value such as +1 or −1 to the original image. Therefore, in the attack, a slightly larger change than this change is given. The data was erased due to small image quality degradation, and the embedded information could not be detected accurately. On the other hand, when the fixed value of embedding is increased, there is a problem that the image quality deteriorates and cannot be used.

  Also, since the magnitude of the attack is unknown, there is a problem of how much embedding strength is set.

  Further, if one digital watermarking system knows the method, the attack method against it will be researched and broken, and there is a problem that it is not possible to detect normally though it is desired to superimpose a plurality of digital watermark elements.

  The present invention has been made to solve the above-described problems, and its purpose is to guarantee the image quality after embedding a digital watermark, information remains even in the case of a large image quality degradation attack, and the size of the attack is small. Since it is unknown, the maximum value of change can be made sufficiently large, data can be embedded as much as possible to improve detection accuracy, and digital watermark elements can be superimposed to accurately detect each embedded information A digital watermarking system that can be obtained is obtained.

  A digital watermark system according to the present invention includes a digital watermark embedder that embeds watermark information in a first image, and a digital watermark detector that detects watermark information from a second image in which the watermark information is embedded by the digital watermark embedder. An electronic watermarking system, wherein the electronic watermark embedder performs error correction coding by repeating each information bit for the watermark information to make redundant data, and obtains embedded data; An embedding position determiner that determines an embedding position based on a key, an embedding distribution function generator that generates an embedding absolute value that changes for each embedding data, and an embedding pattern based on the embedding data, the embedding position, and the embedding absolute value. And an embedded pattern generator for generating the embedded pattern in the first image. Those having an adder for adding.

  The digital watermark system according to the present invention has an effect that it is possible to quantitatively set image quality deterioration and attack resistance after embedding watermark information.

Embodiment 1 FIG.
A digital watermark system according to Embodiment 1 of the present invention will be described with reference to FIGS. FIG. 1 is a block diagram showing a configuration of a digital watermark system according to Embodiment 1 of the present invention. In addition, in each figure, the same code | symbol shows the same or equivalent part.

  In FIG. 1, the digital watermark system according to Embodiment 1 of the present invention is provided with a digital watermark embedder 100A and a digital watermark detector 200.

  The digital watermark embedder 100A is provided with an error correction encoder 101, an embedding position determiner 102, an embedding pattern generator 103, an adder 104, and an embedding distribution function generator 105.

  The digital watermark detector 200 is provided with a subtractor 201, a detection data generator 202, and an error correction decoder 203.

  FIG. 2 is a block diagram showing the configuration of the embedded distribution function generator of the digital watermark system according to Embodiment 1 of the present invention.

  In FIG. 2, the embedded distribution function generator 105 includes a random number generator 1051, a plurality (12) of numerical registers 1052, and a comparator 1053.

  In the first embodiment, the image quality is guaranteed by defining the distribution of changes in the digital watermark. Also, by increasing the distribution variance, it is possible to accurately detect the embedded information because part of the error correction code remains even against a strong attack.

  Next, the operation of the digital watermark system according to the first embodiment will be described with reference to the drawings.

  FIGS. 3A to 3C are diagrams showing the distribution of the change amount of each pixel after the digital watermark data is embedded by the digital watermark system according to Embodiment 1 of the present invention.

  The embedding distribution function generator 105 defines the embedding size distribution of the digital watermark and generates an embedding absolute value (105d) E that changes for each embedding data 101d. The embedding pattern generator 103 generates an embedding pattern 103d as + E when the embedding data 101d is “1” and −E when it is “0”.

  That is, the random number generator 1051 generates a real random number from 0 to 1. The comparator 1053 compares this random number with the m number registers 1052 having sequentially increasing values from 0 to 1, and the first number E (0 to m) at which the random number is larger than the number register 1052. Is embedded and output as an absolute value 105d. However, when it is smaller than the numerical register 1, E = 0 is set, indicating that no data is embedded.

  The numerical value of the register corresponds to, for example, the medium distribution function of the change amount in FIG. 3B, the distribution of 0 is 0.790, and the distribution of ± 1 to ± 5 other than 0 is 0.042. Yes, register 1 is 0.790, registers 2-5 are numerical values obtained by sequentially adding the occurrence distribution 0.042, and register 6 and later are 1.000. Note that three numerical values corresponding to the distribution functions shown in FIGS. 3A to 3C may be set in advance in each register 1052, and one of them may be selected.

  In FIG. 3, (a) shows a distribution function with a small amount of change, (b) shows a distribution function with a medium amount of change, and (c) shows a distribution function with a large amount of change. The distribution function with a small change amount shown in (a) has a small change amount from −3 to +3, but a large proportion of distributions other than 0. The distribution function having a large change amount shown in (c) has a large change amount from −12 to +12, but the distribution ratio other than 0 is small. Further, the medium distribution function of the change amount shown in (b) is intermediate, and these image quality degradations are equal at S / N = -3 dB, and are visually equivalent.

  The distribution function (a) with a small amount of change is a weak attack in which ± 2 is added at random. Most of the data is erased, and a moderate attack in which ± 4 is added randomly and a strong attack in which ± 6 is added randomly. It is completely erased by the attack.

  In the distribution function (c) having a large change amount, most of the data remains in a weak attack, and a large amount of data remains in a moderate attack and a strong attack, so that information embedded by error correction can be accurately detected. .

  The above is a description of a simple digital watermarking method in which a numerical value is added directly to a pixel value of an image. For example, DCT is performed by dividing an image into blocks, and a prescribed change is given to frequency domain components. It may be defined by image quality degradation at the time, and the same applies even if the type of digital watermark is different.

  In the digital watermark system according to the first embodiment, in a digital watermark in which a large number of data including redundant bits for error correction of information to be embedded in an image is embedded, the distribution of the change amount of each pixel after the data is embedded is defined and distributed. In addition, even when a strong attack with a large image quality deterioration occurs, a part of data with a large change amount remains, so that information embedded by error correction can be detected from the remaining data.

  In other words, image quality can be ensured by defining the distribution of changes in the digital watermark. Further, by increasing the distribution variance, it is possible to accurately detect the embedded information because a part of the error correction code remains even against a strong attack. Image quality degradation and attack resistance after embedding watermark information can be set quantitatively.

Embodiment 2. FIG.
A digital watermark system according to Embodiment 2 of the present invention will be described with reference to FIG. FIG. 4 is a diagram showing a distribution of change amounts of each pixel after embedding digital watermark data by the digital watermark system according to the second embodiment of the present invention. The configuration of the digital watermark system according to the second embodiment of the present invention is the same as that of the first embodiment.

  In the second embodiment, by reducing the distribution of the region having a large change, the average image quality deterioration of the image embedded with the digital watermark data is within the specified value even if the maximum value of the change is sufficiently large. is there.

  When the image quality degradation during digital watermark embedding is less than the specified value, the strength of the attack determined by the attacker is unknown, so a large distribution of changes is required. The specified value will be exceeded. Therefore, by setting the distribution function as shown in FIG. 4 in the numerical value register 1052 of the embedded distribution function generator 105, the average image quality degradation of the image is reduced to a predetermined value or less by the distribution function in which the distribution decreases according to the amount of change. This distribution allows a sufficiently large distribution.

  The digital watermarking system according to the second embodiment changes the embedding strength distribution so as to gradually decrease in accordance with the amount of change, and even if the maximum amount of change is increased to a certain value, This guarantees image quality degradation below a specified value.

  In other words, by reducing the distribution of regions with large changes, even if the maximum value of the change is sufficiently large, the average image quality degradation of the image embedded with the digital watermark data can be kept within a specified value. Even against a strong attack, it is possible to obtain a distribution of the amount of change in embedding in which error correction data remains and information can be detected.

Embodiment 3 FIG.
A digital watermark system according to Embodiment 3 of the present invention will be described with reference to FIGS. FIG. 5 is a block diagram showing a configuration of a digital watermark system according to Embodiment 3 of the present invention.

  In FIG. 5, an electronic watermark embedder 100B includes an error correction encoder 101, an embedded position determiner 102, an embedded pattern generator 103, an adder 104, an embedded distribution function generator 105, and an image feature detector. 106. Other configurations are the same as those in the first embodiment.

  In each register 1052 of the embedded distribution function generator 105, three numerical values corresponding to the distribution functions shown in FIGS. 6A to 6C are set in advance, and one of them is selectively output according to the feature of the image.

  Next, the operation of the digital watermark system according to the third embodiment will be described with reference to the drawings.

  FIGS. 6A to 6C are diagrams showing distributions of change amounts of respective pixels after embedding digital watermark data by the digital watermark system according to Embodiment 3 of the present invention.

  In the third embodiment, several distribution functions having different embedding amounts are prepared, and a large distribution function is assigned to a region that is not conspicuous even if the watermark is strongly embedded, such as a portion where the amount of change in the image is large.

  The image feature detector 106 distinguishes a region that is not conspicuous even if the watermark is strongly embedded, such as a portion where the change amount of the image of the original image 10 is large, and the embedded distribution function generator 105 changes the distribution for each region.

  In FIG. 6, the distribution function with a small embedding amount shown in FIG. 6A has an image quality degradation of S / N = −3 dB, and the distribution function with a medium embedding amount shown in FIG. The distribution function with a large embedding amount shown in −6 dB and (c) has an image quality degradation of S / N = −9 dB.

  The embedding area is emphasized in the distribution function (a) with a small embedding amount in the embedding area. The embedding area is inconspicuous in the distribution function (c) with the embedding amount in the non-conspicuous area. Data is embedded according to the medium quantity distribution function (b).

  An image including a region according to the distribution function (c) having a large embedding amount has a large amount of remaining data even when subjected to a strong attack, and at the time of detection, accurate information with high reliability can be detected by error correction decoding.

  In the digital watermark system according to the third embodiment, the distribution of the embedding strength is changed for each embedding position according to the characteristics of the original image 10.

  In other words, prepare several distribution functions with different embedding amounts, assign large distribution functions to inconspicuous areas even if the watermark is strongly embedded, such as areas where the amount of change in the image is large, and embed it strongly in areas that are difficult to see, increasing detection accuracy can do.

Embodiment 4 FIG.
A digital watermark system according to Embodiment 4 of the present invention will be described with reference to FIG. FIG. 7 is a block diagram showing a configuration of a digital watermark system according to Embodiment 4 of the present invention.

  In FIG. 7, a digital watermark system according to Embodiment 4 of the present invention includes a plurality of digital watermark embedders (elements) 100X, which is one of the digital watermark embedders of the above first to third embodiments, A plurality of digital watermark detectors (elements) 200X, which is one of the digital watermark detectors of the first to third embodiments, are provided.

  Further, 1X is each element embedding information, and 3X is each element detection information. That is, each of the plurality of digital watermark embedders (elements) 100X includes the embedded distribution function generator 105 and the like. Other configurations are the same as those in the first embodiment.

  Next, the operation of the digital watermark system according to the fourth embodiment will be described with reference to the drawings.

  In the fourth embodiment, digital watermark elements that guarantee small image quality deterioration of embedding and that can accurately detect information and that guarantee maximum image quality deterioration of an attack are sequentially embedded in a column and embedded. In this case, each embedded information is detected.

  On the digital watermark embedding side, each element embedding information i (1X) (i = 1, 2,..., K) is sequentially embedded in the original image 10 by the digital watermark embedder element i (100X). A rear image 11 is obtained.

  The attacker attacks the post-embedding image 11 with the attacking device 12 to obtain the post-attack image 13.

  On the digital watermark detection side, each element detection information i (3X) is obtained from the post-attack image 13 directly by the digital watermark detector element i (200X) regardless of the embedding order.

When the digital watermark elements 1 to k are superimposed in the order of 1 to k, the image quality degradation when embedding each element is N1 to Nk, the image quality degradation of the attack guaranteed by each element is A1 to Ak, and the embedding made after the superimposition When the image quality degradation of the attack on the rear image 11 is A,
A1> N2 + N3 +... + Nk + A,
A2> N3 + N4 +... + Nk + A,
A3> N4 + N5 +... + Nk + A,
A (k−1)> Nk + A,
Ak> A
Since the element detection information i (3X) satisfies the requirement for accurate detection by superimposing so as to satisfy the above, all can be detected accurately.

  The digital watermark system according to the fourth embodiment guarantees correct information detection against an attack that is equal to or less than the attack image quality degradation prescribed value, and has a plurality of image quality degradations that are less than the embedded image quality degradation prescribed value that is smaller than the attack image degradation prescribed value. In the digital watermark system that sequentially superimposes the digital watermark elements, the information of each watermark element is all overlaid by superimposing it after the attack of each watermark element so as to be larger than the sum of the image quality degradation and the attack of the watermark to be superimposed. This makes it possible to detect accurately.

  In other words, it is guaranteed that there is little degradation in the image quality of embedding, and watermark elements that guarantee the maximum image quality degradation of attacks that can accurately detect information are embedded side by side in series, and each embedded information is embedded in each watermark element from the post-attack image. Can be detected. Information of each digital watermark element can be accurately detected by superimposing a plurality of digital watermark elements.

It is a block diagram which shows the structure of the digital watermark system which concerns on Embodiment 1 of this invention. It is a block diagram which shows the structure of the embedding distribution function generator of the digital watermark system which concerns on Embodiment 1 of this invention. It is a figure which shows distribution of the variation | change_quantity of each pixel after the digital watermark data embedding by the digital watermark system based on Embodiment 1 of this invention. It is a figure which shows distribution of the variation | change_quantity of each pixel after the digital watermark data embedding by the digital watermark system based on Embodiment 2 of this invention. It is a block diagram which shows the structure of the digital watermark system which concerns on Embodiment 3 of this invention. It is a figure which shows distribution of the variation | change_quantity of each pixel after the digital watermark data embedding by the digital watermark system which concerns on Embodiment 3 of this invention. It is a block diagram which shows the structure of the digital watermark system which concerns on Embodiment 4 of this invention. It is a block diagram which shows the structure of the conventional digital watermark system. It is a figure which shows distribution of the variation | change_quantity of each pixel after the data embedding of the digital watermark by the conventional digital watermark system.

Explanation of symbols

  1 embedded information, 2 keys, 3 detection information, 10 original image, 11 post-embedding image, 12 attack device, 13 post-attack image, 100A, 100B digital watermark embedding device, 101 error correction encoder, 102 embedding position determination device, 103 Embedded pattern generator, 104 adder, 105 embedded distribution function generator, 106 image feature detector, 200 digital watermark detector, 201 subtractor, 202 detection data generator, 203 error correction decoder, 1051 random number generator, 1052 Numeric register, 1053 comparator.

Claims (4)

A digital watermark embedder for embedding watermark information in the first image;
A digital watermark system comprising: a digital watermark detector for detecting watermark information from a second image in which watermark information is embedded by the digital watermark embedder;
The digital watermark embedder is:
An error correction encoder that performs embedded error correction coding by repeating each information bit for the watermark information to obtain redundant data; and
An embedding position determiner that determines an embedding position based on a key;
An embedded distribution function generator that generates an embedded absolute value that changes for each embedded data;
An embedding pattern generator for generating an embedding pattern based on the embedding data, the embedding position and the embedding absolute value;
An electronic watermarking system comprising: an adder that adds the embedded pattern to the first image. 2. The digital watermark according to claim 1, wherein the embedded distribution function generator generates an embedded absolute value in which the distribution of the change amount of each pixel after the watermark information is embedded gradually decreases according to the magnitude of the change amount. system. The digital watermark embedder is:
An image feature detector for distinguishing regions based on the amount of change in the first image;
The digital watermarking system according to claim 1, wherein the embedded distribution function generator generates an embedded absolute value in which a distribution of change amounts of each pixel is different for each region. An error correction coder that obtains embedded data by performing error correction coding to make each information bit redundant data for watermark information, an embedded position determiner that determines an embedded position based on a key, and changes for each embedded data An embedding distribution function generator for generating an embedding absolute value, an embedding pattern generator for generating an embedding pattern based on the embedding data, the embedding position and the embedding absolute value, and adding the embedding pattern to the first image There is an adder, and the image quality degradation of the attack guaranteed by each watermark information is larger than the sum of the image quality degradation when embedding each watermark information and the image quality degradation of the attack on the post-embedding image formed after superimposition. In this way, a plurality of electronic images that are sequentially superimposed and embedded with a plurality of watermark information on the first image And lyrics embedded device,
A digital watermark system comprising: a plurality of digital watermark detectors respectively detecting a plurality of watermark information from a second image in which the plurality of watermark information are sequentially superimposed and embedded by the plurality of digital watermark embedders; .

Images