JP3712696B2 - Digital watermark embedding apparatus and program thereof - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention reads an image content to be embedded with an electronic watermark, and embeds the electronic watermark in all blocks constituting the image content using an appropriate method based on the type and the embedding strength. And the program.
[0002]
[Prior art]
Digital watermarks are used to prevent unauthorized use of content data and to manage copyrights by processing content data such as images that cannot be seen with the naked eye.
For example, a typical method is to decompose an image into frequency components and add a correction signal only to a specific band. Although there is no sense of incongruity with the naked eye, various information is embedded by analyzing with corresponding software. In addition, this digital watermark is effective even after processing such as trimming and enlargement / reduction, and has attracted attention because it can be widely countered against illegal conversion of images that has begun to stand out with the generalization of WWW (World Wide Web).
[0003]
Conventionally, in order to apparently avoid image quality degradation after embedding a digital watermark, it is perceived by humans using a masking effect (a visual effect that a weak signal near a strong signal is difficult to perceive by humans). A lot of information was embedded in the difficult high frequency part. This is because E.I. T.A. Lin and E.M. J. et al. Delp, "A Review of Data Hiding in Digital Images," Processeds of the Image Processing, Image Quality Systems, Image Capture Systems Conference, PICS. '99, April 25-28, 1999, Savannah, Georgia (Reference 1).
Here, the “high-frequency portion” refers to a region where the luminance changes greatly between adjacent pixels. On the other hand, a region where the change is minute is called a “flat portion”.
[0004]
[Problems to be solved by the invention]
However, according to the above-described conventional technology, since control of the embedding strength in the flat portion or the contour of the subject is not taken into consideration, there is a possibility that noise occurs in the flat portion or the contour.
In addition, when a person recognizes an object, it uses the visual characteristic of paying particular attention to the contour, and each block of the image divided into blocks is distinguished into a contour and a flat portion, and digital watermarking is performed according to the distinction. By changing the strength of embedding, the image quality deterioration after embedding a digital watermark was suppressed. This is disclosed in Japanese Patent Application Laid-Open No. 2001-8019 (Document 2), Isao Echizen, Yutaka Yoshiura, Kosuke Anjo, Junichi Taguchi, Hiroshi Kurosu, Ryoichi Sasaki, Satoru Tezuka, [Digital watermark image quality maintenance method based on contour preservation] IPSJ Transactions, Vol. 41, no. 6, June, 2000. (Reference 3).
However, in this case, since control of the embedding strength in the attention portion is not taken into consideration, noise occurs in the attention portion and there is a possibility that deterioration of image quality is easily perceived.
[0005]
Although it is possible to detect the high-frequency portion, the flat portion, and the contour using the techniques disclosed in the above-mentioned literature 2 and literature 3, it is generally difficult to accurately detect the attention portion with these techniques. It is. Here, the “consideration part” refers to a conspicuous part on the outline of a thing, for example, a part where a corner, a line segment, or a curve is connected, which is recognized when a person recognizes the shape of the thing.
Therefore, in order to detect a high-frequency part, a flat part, and an outline, the conventional methods shown in Documents 2 and 3 are used, and in order to detect a noticed part, for example, C.I. Harris and M.M. The conventional method shown in Stephens, “A combined corner and edge detector”, Alvey Vision Conference, pp 147-151, 1988 (Document 4) must be used, and it is very useful from the viewpoint of throughput, processing time, and cost. I can't say that.
Moreover, in the above-described conventional method (Document 4), many places that cannot be said to be a noticed part even if viewed by humans are detected as a noticed part.
[0006]
On the other hand, Hiroyuki Tanaka and Masaomi Nakajima “Frequency Transform Digital Watermarking Strengthening Method for Geometrical Modification”, Information Processing Research Report, vol. 98, No. 108, Nov, 1998 (Reference 5) In addition, since the marker used in the digital watermark has conventionally used the pixel bit pattern, the bit pattern of the pixel value changes when the code compression such as JPEG or MPEG is applied by modification such as enlargement / reduction or rotation. Therefore, there is a problem that the marker cannot be specified, and as a result, the embedded information cannot be read.
[0007]
The present invention has been made in view of the various circumstances described above, and classifies each block of an image to be embedded with a digital watermark into four types of a high-frequency part, a contour part, a flat part, and a noticed part. An object of the present invention is to provide a digital watermark embedding device and a program for making it difficult to perceive degradation of image quality due to embedding of the image.
In addition, by detecting the attention part accurately using a method for detecting the attention part from the image, and using the information of the detected attention part, the high frequency part, the flat part, the contour part, and the attention part are determined. It is another object of the present invention to provide a digital watermark embedding device and a program thereof that prevent an increase in processing amount, time, and cost.
[0008]
Furthermore, by applying discriminant analysis to the attention portion obtained based on the technique disclosed in Document 4, the attention portion is determined so as to be appropriate for human beings, and at the same time, the high frequency portion, the flat portion, Another object of the present invention is to provide a digital watermark embedding device and its program that can also classify contour portions and learn the discriminant function as necessary to improve accuracy.
In addition, when an image with an embedded digital watermark is subjected to geometric modification such as enlargement / reduction, rotation, etc., and compression by encoding such as JPEG or MPEG, an attention point such as a corner of the image (interest point) Another object of the present invention is to provide a digital watermark embedding device and a program thereof that can specify a marker by utilizing its high resistance to modification.
[0009]
[Means for Solving the Problems]
In order to solve the above-described problems, the present invention reads out image content to be embedded with a digital watermark, and uses all appropriate blocks based on the type and embedding strength for all the blocks constituting the image content. An electronic watermark embedding apparatus that embeds a watermark, wherein an image block that is an embedding unit of the electronic watermark is adjacent to a high-frequency portion that is a region in which luminance changes greatly between adjacent pixels due to the characteristics of the pixels in the block. Divide into at least four types of attention parts where the image including the flat part, contour part, corner and vertex, line segment and curve connection part where the luminance change between the pixels is small changes significantly in two dimensions, An attention part is detected from the image block, and the high-frequency part, the flat part, the contour part, and the attention part are detected using information on the detected attention part. Min identifying means for determining a digital watermark embedding intensity in each, characterized by comprising a.
[0010]
According to the present invention, the identification unit classifies the image block, which is a digital watermark embedding unit, into at least four types of a high-frequency part, a flat part, a contour part, and a noticed part according to the characteristics of the pixels in the block. By detecting the attention part from the block and determining the digital watermark embedding strength in each of the high-frequency part, flat part, contour part, and attention part using the information of the detected attention part, the image quality of the embedded watermark is improved. It is possible to provide a digital watermark embedding device that makes deterioration difficult to perceive and prevents increase in processing amount, time, and cost.
Note that the “high frequency part” is an area where the luminance change between adjacent pixels is severe, the “flat part” is an area where the luminance change between adjacent pixels is small, and the “attention part” is a corner or vertex. Each of the regions where the image including the line segment and the connecting portion of the curve changes significantly two-dimensionally.
[0011]
Further, in the present invention, the attention part is detected by using the determinant and trace of the averaged autocorrelation matrix at each point of the image to be embedded with the digital watermark, and the strength of the candidate point of the interest point for each point. And an interest point that is larger than a threshold value set in advance for each appropriate region is determined.
[0012]
According to the present invention, by using an interest point present in an image to be embedded with a digital watermark as a marker indicating the position of the container, the marker can be specified by geometric modification such as enlargement, reduction, and rotation. Inconveniences such as disappearance can be eliminated, and an effect is also obtained for compression by encoding such as JPEG (Joint / Photographic / Expert / Group) or MPEG (Moving / Picture / Expert / Group).
The “interest point” is a general term for a portion where the image signal changes significantly in two dimensions, such as a corner, a vertex, a line segment, or a connected portion of a curve. C. Schmid, R.M. Mohr, and C.M. Bauckhage, “Evaluation of Interest Point Detectors,” IJComp. Details of interest points are shown in Vis, 37 (2), 151-172, 2000 (reference 5), and it is shown that the reproducibility with respect to geometric modification is high.
[0013]
Further, in the present invention, the determination of the embedding strength is based on statistical information on the strength of the attention portion calculated for each block, for example, where the average value of the strength belongs within some predetermined range. By determining the embedding strength, the embedding strength is determined for each block.
[0014]
In order to solve the above-described problems, the present invention reads out image content to be embedded with a digital watermark, and uses all appropriate blocks based on the type and embedding strength for all the blocks constituting the image content. An electronic watermark embedding program used in an electronic watermark embedding apparatus for embedding a watermark, wherein an image block which is an embedding unit of the electronic watermark is an area in which a luminance change between adjacent pixels is severe due to the characteristics of the pixels in the block. At least four of the attention portions where the image including a certain high-frequency portion, a flat portion that is a region with little change in luminance between adjacent pixels, a contour portion, corners and vertices, line segments and curved line connection portions, and two-dimensionally changes significantly A first step of classifying into two types, and a noticed part is detected from the image block, and the detected noticed part The high frequency part with the information, characterized in that to execute the flat portion, the contour portion, a second step of determining a digital watermark embedding intensity at each noted portion, to the computer.
[0015]
In the present invention, the second step divides the image into blocks of an arbitrary size, and sets a set of statistics such as average and variance of the intensity of the attention portion in the block unit. A step of discriminating each block into a high-frequency part, a flat part, a contour part, and a noticed part using an identification tool composed of a discriminant function group learned in advance by discriminant analysis, and for each block A step of determining the embedding strength for each block by determining where, for example, the average value of the strength belongs within some predetermined range based on the calculated statistical information on the strength of the attention portion. And causing a computer to execute the steps described above.
[0016]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 is a functional block diagram showing the internal configuration of the digital watermark embedding apparatus of the present invention.
The digital watermark embedding apparatus according to the present invention includes an image input unit 11, an image processing unit 12, an identification unit 13, an information embedding unit 14, and an image output unit 15.
[0017]
The image input unit 11 selects an image to be processed from an image storage unit (not shown), extracts luminance data from the image, and passes the luminance data to the image processing unit 12. The image processing unit 12 has a function of calculating the intensity for each point by performing a calculation based on the luminance data obtained from the image input unit 11.
The identification unit 13 has a function of extracting feature amounts in units of blocks, identifying the blocks according to four categories described later, and calculating an embedding strength in units of blocks. Specifically, an image block, which is a digital watermark embedding unit, has a high-frequency portion, which is a region where the luminance change between adjacent pixels is sharp, due to the characteristics of the pixels in the block, and the luminance change between adjacent pixels is small. The image including the flat part, outline part, corners and vertices, line segments and curved lines, which are areas, is classified into at least four types of attention parts that change significantly in two dimensions, and the attention part is detected from these image blocks Then, using the information of the detected attention portion, the digital watermark embedding strength in each of the high frequency portion, the flat portion, the contour portion, and the attention portion is determined. Details will be described later.
[0018]
The information embedding unit 14 has a function of embedding a digital watermark using an appropriate method such as a sample value direct operation type or a frequency component operation type based on the type and embedding strength value for all blocks. .
The image output unit 15 outputs an image in which information is embedded to a display or an image storage unit (not shown).
[0019]
2 and 3 are flow charts cited for explaining the operation of the embodiment of the present invention shown in FIG.
The operation of the embodiment of the present invention will be described in detail below with reference to the flowcharts shown in FIGS.
[0020]
FIG. 2 shows a procedure for determining the embedding strength of the digital watermark using the corner as an example of the part to be noted.
In FIG. 2, the image input unit 11 first selects an image to be processed from an image storage unit (not shown), extracts luminance data from the image, and delivers the luminance data to the image processing unit 12 (S21). ). The image processing unit 12 calculates primary differential coefficients of the luminance data obtained through the image input unit 11 in the X direction and the Y direction (S22), and the primary differential coefficients in the X direction and the Y direction at each point of the image. Then, an autocorrelation matrix is obtained from the set (x, y) and an averaging process is performed within an appropriate region including the point of interest to obtain an averaged autocorrelation matrix (S23). Then, the embedding strength of the digital watermark is calculated for each point using the determinant and trace of the averaged autocorrelation matrix at each point of the image (S24). Here, since the attention part is illustrated, the intensity of the corner is calculated, and the result of the image processing as described above is delivered to the identification unit 13.
[0021]
Next, the identification unit 13 divides the image into blocks of an appropriate size, and calculates a set of average and variance of the intensity of the corners as a block feature amount (S25). Then, each block is identified as a high-frequency block, a flat block, a contour block, and a corner block by using an identification tool composed of a discriminant function group previously learned by discriminant analysis (S26).
Here, the embedding strength is determined for each block by determining where the average value of the corner strengths calculated for each block belongs in some predetermined ranges (S27). That is, by performing a comparison check with a threshold range prepared in advance, for example, the embedding strength of the digital watermark is classified into large, medium, and small. Then, for all the blocks via the information embedding unit 14, based on the type (high frequency, flat, contour, corner) and embedding strength value, the sample value direct manipulation type digital watermark embedding or the frequency component manipulation type digital watermark is performed. The digital watermark information is embedded by any method such as embedding (S28).
[0022]
The above operation is continuously performed on all blocks constituting the image, and is output to the display or the image storage unit via the image output unit 15.
[0023]
The “sample value direct manipulation type” described above is a method of embedding information by manipulating sampled information such as a luminance value and an RGB value for each pixel. In order to prevent the presence of a digital watermark from being perceived, knowledge about human visual and perceptual characteristics such as a masking effect is used for selecting an information embedding position or the like. Alternatively, a method is used in which information is distributed and embedded at random positions and is detected using a statistical bias at the time of decoding. In general, compared to the frequency component operation type digital watermarking method, the resistance to geometrical modifications such as enlargement / reduction and partial cutout is excellent, but the deterioration of image quality tends to increase.
The above-mentioned “frequency component operation type” is obtained from the sample value (luminance value, RGB value, etc.) of the image into which information is embedded by orthogonal transform such as Fourier transform, discrete cosine transform, and wavelet transform. The information is embedded by manipulating the frequency component values. The embedded frequency component is returned to the image state by inverse transformation to generate watermarked content. In general, compared to the sample value direct manipulation type digital watermarking method, it is more resistant to non-geometric modification such as gradation conversion and lossy compression, and image quality degradation can be suppressed to a small extent. Therefore, there is a drawback that processing for embedding and extraction takes time.
Both the sample value direct manipulation type and the frequency component manipulation type are known digital watermark embedding techniques.
[0024]
As an example of information embedding using interest points, FIG. 3 shows a digital watermark embedding procedure using the technique of the above-mentioned document 4 and document 5.
Here, as in FIG. 2, the image input unit 11 selects an image to be processed from an image storage unit (not shown), extracts luminance data from the image, and passes the luminance data to the image processing unit 12 (S31). Then, the image processing unit 12 calculates primary differential coefficients of the luminance data obtained via the image input unit 11 in the X direction and the Y direction, respectively (S32), and at each point of the image, the primary direction in the X direction and the Y direction is calculated. An autocorrelation matrix is obtained from the set of differential coefficients (x, y), an averaging process is performed within an appropriate region including the point of interest, and an averaged autocorrelation matrix is obtained (S33).
[0025]
Then, using the determinant and trace of the averaged autocorrelation matrix at each point of the image, the strength of the candidate point of interest point is calculated for each point, and the value larger than the threshold set in advance for each appropriate region is calculated. The point is determined (S34).
Then, digital watermark information is embedded in all the blocks via the information embedding unit 14 by any method such as sample value direct manipulation type digital watermark embedding or frequency component manipulation type digital watermark embedding (S35).
The above operation is continuously performed on all the blocks constituting the image, and is output to a display (not shown) or an image storage unit via the image output unit 15.
[0026]
As described above, by using an interest point in an image to be embedded with a digital watermark as a marker indicating the position of the container, the marker cannot be specified by geometric modification such as enlargement, reduction, or rotation. Such inconveniences can be solved. The fact that interest points have reproducibility for geometric modification is Schmid, R.M. Mohr, and C.M. It is shown in Bauchage, “Comparing and Evaluating Interest Points, International Conference on ComputerVision, Bombay, January, 1998. (Reference 6). By using this property, compression by encoding such as JPEG and MPEG is also used. An effect is obtained.
[0027]
The programs for realizing the functions of the image input unit 11, the image processing unit 12, the identification unit 13, the information embedding unit 14, and the image output unit 15 shown in FIG. The digital watermark embedding apparatus of the present invention is constructed by recording and reading the program recorded on the recording medium into a computer, and the computer sequentially reading and executing the program. The computer system here includes an OS and hardware such as peripheral devices.
[0028]
Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW system is used.
The “computer-readable recording medium” refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, and a CD-ROM, and a storage device such as a hard disk built in the computer system. Further, the “computer-readable recording medium” refers to a volatile memory (RAM) in a computer system that becomes a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In addition, those holding programs for a certain period of time are also included.
[0029]
The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line.
The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.
[0030]
The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes designs and the like that do not depart from the gist of the present invention.
[0031]
【The invention's effect】
As described above, according to the present invention, the human visual characteristics are used to identify a portion where the human being pays particular attention in the image, and the embedding strength of information by digital watermark is weakened in such a place, Since it is possible to make a strong decision on a portion that does not pay much attention, image quality deterioration due to embedding information with a digital watermark is less perceivable than in the conventional method.
As a result, the embedding of identification information and attribute information into digital content using digital watermarks, which had been often avoided in the past due to the risk of image quality degradation, was further promoted. The creation of a content distribution market can also be promoted.
In addition, it only applies a guideline regarding the strength of embedding information in an image using digital watermark, and it does not matter what type of digital watermark method is used for embedding, so it is applicable not only to digital watermark but also to a wide range of information embedding It is possible.
[0032]
Furthermore, according to the present invention, the position of the container can be accurately determined even if the image in which the digital watermark is embedded is subjected to geometric modification such as enlargement, reduction or rotation, and compression by encoding such as JPEG or MPEG. Therefore, even when such an attack is performed, the accuracy of reading the embedded information is improved.
As a result, embedding of identification information and attribute information into digital contents by digital watermarks, which has been conventionally avoided from the viewpoint of vulnerability to the above attack, is promoted, and further, copyright protection and processing are possible. To promote the creation of a secondary market for digital content.
[Brief description of the drawings]
FIG. 1 is a block diagram showing an expanded function of an internal configuration of a digital watermark embedding apparatus according to the present invention.
FIG. 2 is a flowchart cited for explaining the operation of the embodiment of the present invention.
FIG. 3 is a flowchart cited for explaining the operation of the embodiment of the present invention.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 11 ... Image input part, 12 ... Image processing part, 13 ... Identification part, 14 ... Information embedding part, 15 ... Image output part

Claims (5)

An electronic watermark embedding device that reads out image content to be embedded with an electronic watermark, and embeds the electronic watermark into all blocks constituting the image content using an appropriate method based on the type and the embedding strength,
The image block, which is an embedding unit of the digital watermark, is flat according to the characteristics of the pixels in the block, a high frequency portion where the luminance change between adjacent pixels is severe, and a flat region where the luminance change between adjacent pixels is small The image including the part, contour part, corner and vertex, line segment and curve connection part is divided into at least four types of attention parts that change significantly in two dimensions, and the attention part is detected from the image block, and the detection is performed. Identification means for determining a digital watermark embedding strength in each of the high-frequency portion, the flat portion, the contour portion, and the attention portion using the information on the attention portion
An electronic watermark embedding device comprising: Detection of the attention part is as follows:
Calculating the strength of interest point candidate points for each point using the determinant and trace of the averaged autocorrelation matrix at each point of the image to be embedded with the digital watermark, and a threshold set in advance for each appropriate region Determine an interest point that is greater than
The digital watermark embedding apparatus according to claim 1. The determination of the embedding strength is as follows:
Based on the statistical information of the intensity of the attention portion calculated for each block, for example, by determining where the average value of the intensity belongs within some predetermined range, the embedding strength can be set for the block unit. The digital watermark embedding apparatus according to claim 1, wherein the determination is performed. A digital watermark used in a digital watermark embedding apparatus that reads out image content to be embedded with a digital watermark and embeds the digital watermark in all blocks constituting the image content using an appropriate method based on the type and the embedding strength. An embedded program,
The image block, which is an embedding unit of the digital watermark, is flat according to the characteristics of the pixels in the block, a high frequency portion where the luminance change between adjacent pixels is severe, and a flat region where the luminance change between adjacent pixels is small A first step of classifying an image including a portion, a contour portion, a corner and a vertex, a line segment and a curved line connection portion into at least four types of attention portions that are remarkably changed two-dimensionally;
A second step of detecting an attention part from the image block and determining a watermark embedding strength in each of the high-frequency part, the flat part, the contour part, and the attention part using statistical information of the detected attention part;
A watermark embedding program that causes a computer to execute. In the second step, the image is divided into blocks of an arbitrary size, and a set of statistics such as an average intensity and a variance of the attention portion is calculated as a feature amount of the block for each block.
Identifying each block into a high-frequency part, a flat part, a contour part, and an attention part using an identification tool composed of a discriminant function group previously learned by discriminant analysis;
The embedding strength is determined for each block by determining where the average value of the strength belongs within some predetermined range based on the statistical information of the attention portion calculated for each block. Including the steps of:
5. The digital watermark embedding program according to claim 4, which causes a computer to execute each step.

Images