JP3722995B2 - Watermark encoding method and decoding method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a watermark (digital watermark) encoding method and decoding method, and more particularly, to an improvement for reducing the data amount (signal amount) of original image data to be stored.
[0002]
[Prior art]
Watermark technology is technology that prevents secondary use such as illegal copying by embedding copyright and purchaser information in image data, which is data (signal) representing an image, in a form that is not perceived by the user. It is. The distribution source of copyrighted image data stores the image data before embedding the watermark, that is, the original image data, and collects the image data that is suspected of illegal copying from the city. By comparing with the original image data, the watermark is taken out and further decoded into copyright information or the like. The distribution source performs purchaser identification and the like based on the decrypted information.
[0003]
FIG. 10 and FIG. 11 are flowcharts showing processing procedures of a watermark encoding method and a decoding method, respectively, conventionally performed. FIG. 12 is a timing chart showing the transition of the image format in these processes. These processes are executed by a distribution source (or a person authorized by the distribution source).
[0004]
In a conventional watermark encoding method, original image data having an uncompressed original image format is compressed based on an image compression technique such as JPEG. Thereby, the image data is distributed in the form of compressed image data. In this compression process, watermarks are embedded. On the other hand, the original image data described above is stored in the format of the original image.
[0005]
The watermark encoding method will be described with reference to FIG. 10 and FIG. 12. The original image data as the original image is obtained by, for example, spectral data (spatial spectral data) SPC by DCT conversion which is a process of JPEG compression processing. (Step S51). Spectral data includes a number of spectral components as its constituent elements. The watermark embedded with the copyright information and the like is embedded in the spectrum data SPC (step S52). Spectral data SPC is obtained by expanding an original image into a set of spectral components, ie, a set of spatial frequency components, and the DCT transform is a typical example of such expansion.
[0006]
Subsequently, the spectrum data SPC including the watermark is converted into a VLC code (variable length code) through quantization and VLC coding (variable length coding) (step S53). Then, the image data is distributed to the market through a distributor or the like in the VLC code format, that is, the compressed image data format (step S54). The original image data is stored in the distribution source in the original image format (step S55).
[0007]
Next, a conventional watermark decoding method will be described with reference to FIGS. The image data collected in the city may be in the VLC code format or the original image format. Regardless of which format is used, the image data is converted into spectral data SPC ′ (step S61).
[0008]
For example, if the collected image data is in VLC code format, it is converted into spectral data SPC 'through VLC decoding, which is the inverse operation of VLC encoding, and inverse quantization, which is the inverse operation of quantization. The If the collected image data is in the form of an original image, it is converted into spectral data SPC ′ through DCT conversion.
[0009]
Since the spectrum data SPC ′ obtained through step S61 includes a watermark, the value is different from the spectrum data SPC shown in FIG. By calculating the difference between the two, the watermark embedded in the spectrum data SPC ′ is extracted (step S62). Next, the extracted watermark is decrypted (step S63), whereby copyright information and the like are read out.
[0010]
In general, the collected image data may not be the image data immediately after distribution, but may be through various image processes. For this reason, there is a possibility that noise derived from these image processes is mixed in the collected image data. This noise also causes a deviation between the spectrum data SPC in FIG. 10 and the spectrum data SPC ′ in FIG.
[0011]
In order to eliminate the adverse effect of noise in taking out the watermark, in the watermark embedding process (step S52) in FIG. 10, the watermark is repeatedly embedded in one image, and the taking process (step In S62), a method of increasing the S / N ratio by calculating the average of repeated watermarks is employed.
[0012]
Alternatively, an error correction code is included in the watermark in the watermark embedding process (step S52), and the decoding error is detected and corrected in the extraction process (step S62). The method of improving the reliability of is taken. In any method, in step S63, whether or not there is reliability is determined. When it is determined that there is reliability, copyright information and the like are output.
[0013]
[Problems to be solved by the invention]
However, in the conventional method, since the original image data is stored in the format of the original image, the distribution source that normally needs to store a large number of types of original image data has a total data of the original image data to be stored. There was a problem that the amount became large. When the amount of data increases, it is inconvenient in storage and handling, and the necessary cost increases.
[0014]
The present invention has been made to solve the above-described problems in the conventional method, and an object thereof is to provide a watermark encoding method and a decoding method that can reduce the amount of original image data to be stored. And
[0015]
[Means for Solving the Problems]
According to a first aspect of the present invention, there is provided a watermark encoding method for embedding a watermark in original image data, wherein (a) the original image data is not compressed through image compression including conversion into spectral data and quantization. Converting from an image format to a compressed image data format; (b) storing the original image data in the compressed image data format; and (c) inverse quantization as an inverse operation of the quantization. Converting the original image data from the compressed image data format to spectral data through image decompression including: (d) superimposing a watermark on the spectral data obtained in the step (c); (E) converting the spectral data on which the watermark is superimposed in the step (d) into a format of the compressed image data through image compression including the quantization; and (F) and a, a step of delivering the compressed image data obtained by converting in the step (e).
In addition, the step (d) includes (d-1) a step of superimposing random fluctuations on the spectrum data obtained in the step (c), and (d-2) the step (d-1). ) To obtain a quantized coefficient by performing quantization on the spectral data on which random fluctuations are superimposed using a scaling coefficient larger than the scaling coefficient used in the quantization, and (d-3) selecting the embedding target of the watermark from spectral components that constitute the spectral data and give a non-zero value as the quantization coefficient; and (d-4) the step (d- Superimposing the watermark on a value before the fluctuation of the spectral component selected in 3) is superimposed.
[0016]
According to a second aspect of the present invention, there is provided a watermark encoding method for embedding a watermark in original image data, wherein: (a) the original image data is not compressed through image compression including conversion into spectral data and quantization. Converting from an image format to a compressed image data format; (b) storing the original image data in the compressed image data format; and (c) inverse quantization as an inverse operation of the quantization. Converting the original image data from the compressed image data format to spectral data through image decompression including: (d) superimposing a watermark on the spectral data obtained in the step (c); (E) converting the spectral data on which the watermark is superimposed in the step (d) into a format of the compressed image data through image compression including the quantization; and (F) and a, a step of delivering the compressed image data obtained by converting in the step (e).
Moreover, the step (d) includes the step (d-1) of setting the amplitude of the watermark to be superimposed to a value larger than ½ times the quantization step used in the quantization. Yes.
[0017]
According to a third aspect of the present invention, in the watermark encoding method of the second aspect, the step (d) is based on (d-2) a predetermined constant L (≦ 1/2) set in advance. When the amplitude set in the step (d-1) is smaller than L times the value of the quantization table, the amplitude is set to a predetermined constant K (L ≦ K ≦ 1/2) of the value of the quantization table. And a step of setting to a double value.
[0018]
According to a fourth aspect of the present invention, there is provided a watermark decoding method for extracting a watermark from collected image data, wherein the watermark is stored in step (b) of any one of the first to third watermark encoding methods. For the original image data, the step of obtaining spectral data by executing the same step as the step (c), (B) the step of converting the collected image data into spectral data, C) calculating the difference between the two spectral data obtained in the steps (A) and (B), and extracting the watermark from the spectral data obtained in the step (B). I have.
[0019]
DETAILED DESCRIPTION OF THE INVENTION
<1. Outline procedure of encoding method and decoding method>
In order to save the data amount of the original image data to be stored, in the preferred embodiment of the present invention, the original image data is stored not in the original image format but in the VLC code format, that is, the compressed image data format. Is done. FIG. 1 and FIG. 2 are flowcharts showing the processing procedures of the watermark encoding method and decoding method of this embodiment, respectively. FIG. 3 is a timing chart showing the transition of the image format associated with these processes. These processes are executed by a distribution source (or a person authorized by the distribution source). 1 to 3 are respectively compared with FIGS. 10 to 12 showing the conventional method.
[0020]
The watermark coding method according to the embodiment will be described with reference to FIGS. 1 and 3. Original image data as an original image is compressed based on, for example, JPEG, and converted into VLC code, that is, compressed image data. Conversion is performed (step S1). The original image data is stored in this VLC code format (step S2). That is, the original image data is stored in an image format with a small data amount. As a result, the data amount of the original image data to be stored is reduced, and the problems in the conventional watermark encoding method are alleviated or eliminated.
[0021]
In order to embed a watermark in original image data, first, VLC decoding and inverse quantization are performed on the original image data having the VLC code format, and the image is expanded to spectral data SPC through them. (Step S3). Next, a watermark including the encoded copyright information and the like is embedded in the spectrum data SPC (step S4).
[0022]
Subsequently, the spectrum data SPC including the watermark is converted into a VLC code through quantization and VLC encoding (step S5). Then, the image data is distributed to the market through a distributor or the like in the VLC code format, that is, the compressed image data format (step S6).
[0023]
Next, a watermark decoding method corresponding to the method of FIG. 10 will be described with reference to FIGS. The image data collected in the city may be in the VLC code format or the original image format, as in the past. Regardless of which format is used, the image data is converted into spectral data SPC ′ (step S11).
[0024]
Next, VLC decoding and inverse quantization are performed on the original image data in the VLC code format, and spectral data SPC is obtained through them (step S12). This process is performed in the same way as the process in step S3. Since the spectrum data SPC ′ obtained through step S11 includes a watermark, the value is different from the spectrum data SPC obtained in step S12.
[0025]
By calculating the difference between the two, the watermark embedded in the spectrum data SPC ′ is extracted (step S13). Next, the extracted watermark is decrypted (step S14), whereby copyright information and the like are read out. Steps S11, S13, and S14 are performed in the same manner as the conventional steps S61, S62, and S63 (FIG. 11), respectively. For this reason, detailed description of these processes is omitted.
[0026]
<2. Problems to be improved in encoding method>
By the above method, the problem that occurs in practical use that the amount of original image data to be stored is excessive is alleviated or eliminated. However, if the process of embedding the watermark (step S4) is executed in the same form as the conventional step S52 (FIG. 10), the new image quality will be remarkably deteriorated or the decoding accuracy rate will be deteriorated. Problems arise. Here, this problem will be described.
[0027]
The cause of the problem that the image quality deteriorates is in the process of determining the position where the watermark is to be embedded in the image of one screen, and the cause of the problem that the decoding accuracy rate decreases is the watermark to be embedded Exists in the process of determining the amplitude of. Both processes are included in the process of step S4.
[0028]
When compression processing such as JPEG is performed on image data, processing such as DCT conversion, quantization, and VLC code is performed in units of blocks obtained by dividing one screen. The same applies to the reverse operation. Each block is composed of 8 × 8 pixels or 16 × 16 pixels, for example.
[0029]
Among the plurality of blocks, a specific block (generally a plurality) satisfying a certain condition is selected as a position for embedding the watermark, and a specific spectral component in the selected block is selected. The spectral component is a component of the spectral data, and the spectral data is composed of a set of spectral components. As described above, the selected “position for embedding the watermark” includes both a spatial position on the screen and a position in the spectrum.
[0030]
More specifically, for example, four AC components having the lowest spatial frequency (AC low frequency four components) are selected as specific spectral components. Among various spectral components, the low-frequency component has the greatest effect on image quality, so it is difficult to remove watermarks without degrading image quality by image processing such as filtering. is there. That is, this is to increase the strength (difficulty of removal) of the watermark.
[0031]
As the specific block, for example, a block in which the amplitude of each of the above four AC low frequency components is somewhat large and the high frequency component is somewhat large is selected. The condition that the amplitude of the AC4 component is large is set in order to prevent disappearance of the watermark due to quantization after embedding the watermark and to make the watermark difficult to stand out visually. The condition that the high frequency component is large to some extent is to make the watermark more inconspicuous.
[0032]
These conditions per se relating to a specific block and a specific spectral component selected as a position to embed a watermark are well known in the art. Among these three conditions, the condition that the amplitude of the AC4 component is large is more specifically the quantization coefficient obtained when provisional quantization is performed using a larger scaling coefficient SC. Judgment is made by substituting the condition that Q is not zero. “Large” means that it is somewhat larger (for example, by 1) than the scaling factor SC used in the quantization process performed to finally obtain the VLC code.
[0033]
The scaling factor SC is one of the factors that define the quantization step. Image compression algorithms such as JPEG are common to all blocks, but all values belonging to one block are added to the values of the quantization table TB prepared by the algorithm itself so that each spectral component has a separate value. Is multiplied by a scaling factor SC that is common to the spectral components of ## EQU1 ## to give a quantization step [Delta] corresponding to the quantization width (roughness). That is,
Δ ＝ SC × TB ・ ・ ・ ・ (Formula 1)
Given in. The size of the scaling coefficient SC is appropriately set to an appropriate value according to the data amount of the VLC code obtained by compression.
[0034]
As described above, it is determined whether or not the value of the quantization coefficient Q of the AC low frequency 4-component is zero by performing quantization using the scaling coefficient SC that is set to be larger. At this time, the spectrum data SPC to be embedded with the watermark is not obtained by directly converting from the original image, as shown in FIG. 3, but is converted into the VLC code and the subsequent inverse conversion. It was obtained through
[0035]
That is, the spectrum data SPC to be embedded with the watermark is obtained through image compression and image expansion along the path A indicated by the dotted line in FIG. The quantization process is an irreversible process, and even if the quantization and the subsequent inverse quantization are executed using the same quantization step Δ, the spectrum data SPC is the same as the spectrum data SPC before the quantization. It is not played back the same. That is, route A is irreversible.
[0036]
Therefore, for example, when the spectral data SPC obtained by performing the quantization and the inverse quantization along the path A using the scaling coefficient SC = 1 is quantized with the scaling coefficient SC = 2, The relationship between the spectral component and its quantization coefficient Q is drawn with a dotted line in the graph of FIG. In FIG. 4, the horizontal axis X represents the value of the spectral component. The graph of FIG. 4 shows the spectrum when quantization is performed with the scaling factors SC = 1 and SC = 2 on the spectral data SPC obtained directly from the original image without passing through the path A. The relationship between the component X and the quantization coefficient Q is also shown at the same time.
[0037]
As shown in FIG. 4, the quantization coefficient obtained by quantization with the scaling coefficient SC = 2 between the spectrum data SPC once quantized with the scaling coefficient SC = 1 and the spectrum data SPC that is not. Deviations appear in the value of Q. This means that a deviation appears at a position where the watermark is to be embedded between the two spectral data SPC. This deviation causes a problem that the distortion of the image due to the watermark is visually noticeable.
[0038]
The cause of the problem that the accuracy rate decreases is explained by FIG. Since the spectral data SPC to be embedded in the watermark has already undergone the processes of quantization and inverse quantization, each spectral component X is an integer of the quantization step Δ as shown by the white circles in FIG. Can only take double value. After the watermark is embedded, quantization is performed using a quantization step Δ having the same size. Therefore, if the amplitude of the watermark is not properly selected, the watermark will be The mark will not remain. This causes a problem of deterioration of the decoding accuracy rate.
[0039]
In addition, the broken line in FIG. 5 has shown the relationship between the spectrum component X and the quantization coefficient Q which have not passed through the quantization. In the watermark encoding method of the embodiment, as described below, a watermark embedding process (step S4) is executed so as to solve these problems.
[0040]
<3. Watermark embedding process>
FIG. 6 is a flowchart showing an internal flow of the process of embedding a watermark (step S4). In this processing, the luminance component Ysp, the scaling coefficient SC, the watermark WM, and the random number PN of the spectrum data SPC are supplied as processing targets. The luminance component Ysp is one of three elements included in each spectral component that constitutes the spectral data SPC. As the random number PN, a value of +1 or −1 (in other words, a positive or negative sign) is randomly supplied. In the following flowchart, a left-pointing arrow included in a mathematical expression drawn in each step represents that a value on the right side is substituted into a variable on the left side.
[0041]
When the process is started, in step S21, a minute numerical value having a random number PN as a code is assigned to the variable δ. For example,
δ = PN × 10 ^-3 ... (Formula 2)
Defines the variable δ. And as a new variable Ytm
Ytm ＝ Ysp ＋ δ ・ ・ ・ ・ （Formula 3）
Is defined. That is, a fluctuation corresponding to the variable δ is added to the luminance component Ysp, and the variable Ytm is introduced to make this a temporary luminance component.
[0042]
Next, the processes of steps S22 to S26 are repeatedly executed for each block. In step S22, the variable Ytm corresponding to the luminance component Ysp including the fluctuation δ is quantized, and further inverse quantized. That is, the quantization coefficient Q is
Q = round (Ytm / (TB x 2)) (Equation 4)
Calculated based on Then, for this quantization coefficient Q,
Qsp = Q x (TB x 2) ... (Formula 5)
Based on this, inverse quantization is performed to obtain spectral data Qsp as an inverse quantization coefficient.
[0043]
Here, as an example, it is assumed that the scaling coefficient SC used in the quantization performed in steps S1 and S5 in FIG. 1 is SC = 1. For this reason, in Equations (4) and (5), the scaling coefficient SC is set to “2” which is one step larger than “1”. Two variables Q and Qsp obtained by the equations (4) and (5) are used for the determination in the next step S23.
[0044]
The quantization coefficient Q obtained by Expression (4) is a value calculated not based on the luminance component Ysp itself but based on the variable Ytm on which fluctuation (noise) δ is superimposed. Therefore, the relationship between the luminance component Ysp and the quantization coefficient Q is distributed in two ways according to the direction of fluctuation, that is, the sign of the variable δ, as shown in FIG. At the center of dispersion, a relationship (indicated by a one-dot chain line) between the luminance component Ysp obtained by direct conversion from the original image and its quantization coefficient Q without being quantized is located.
[0045]
Therefore, the quantization coefficient Q of the variable Ytm including random fluctuations (random noise) coincides on the average with the quantization coefficient Q of the luminance component Ysp that has not been quantized. For this reason, the above-described problem of image quality degradation is solved.
[0046]
In step S23, the watermark embedding position is determined. An identification number IND is assigned to each determined position. The internal flow of step S23 is represented by the flowchart of FIG. In step S23, the quantization coefficient Q and the spectrum data Qsp are processed, and the identification number IND is output. That is, in step S31, the following three conditions are set as specific spectral components of the specific block to which the identification number IND should be assigned:
Spectral components belong to 4 AC low frequency components (Condition 1)
The quantization factor Q is not zero (Condition 2)
The sum of the absolute values of the AC components of the spectrum data Qsp is a certain number (for example, 200) or more (Condition 3).
[0047]
These conditions specifically express the conditions regarding the specific block and the specific spectrum component described above. When the process of step S31 ends, the selected identification number IND is output, and the process of step S23 is completed. Thereafter, the process proceeds to step S24 in FIG.
[0048]
In step S24, it is determined whether or not the identification number IND has reached a sufficient number for embedding the watermark WM. More specifically, it is determined whether or not two or more bits constituting the watermark are included. If the result of the determination is “No”, a message to that effect is output in step S25. Thereafter, the process may be terminated or continued. If the determination result is “Yes”, the amplitude of the watermark is determined in step S26. The determined amplitude value is assigned to a new variable WMs.
[0049]
FIG. 9 is a flowchart showing the internal flow of step S26. In this process, the scaling factor SC is processed, and the watermark amplitude WMs is output. First, in step S41, it is determined whether or not the scaling coefficient SC is smaller than a value slightly smaller than 1 (for example, 1 / 1.1). If the result of the determination is “Yes”, the value of the temporary variable WMs is
WMs = 0.5 (Equation 6)
Given in. On the other hand, if the result of the determination is “No”, the value of the temporary variable WMs is
WMs = (SC / 2) x 1.1 (Equation 7)
Given in.
[0050]
In step S44, the final variable WMs to be output is
WMs ＝ WMs × TB ・ ・ ・ ・ (Formula 8)
Given in. The equation of Equation 8 is different from the mathematical equation concept, and the equal sign is synonymous with the left-pointing arrow used in the equations in the flowchart.
[0051]
The variable WMs given by Equation 8 is handled as the amplitude of the watermark to be embedded in step S29 described later. That is, the process of step S26 is based on the fact that the spectrum data SPC can take only discontinuous values indicated by white circles in FIG. It means that a value somewhat larger than twice is given. Thus, the fear that the watermark WM disappears due to quantization is eliminated, and the problem of deterioration of the decoding accuracy rate does not occur.
[0052]
It should be noted that if the value of the temporary variable WMs calculated by Equation 7 is 0.5 or less, the distortion of the image resulting from the embedding of the watermark cannot be visually perceived by experience through many simulation experiments. found. For this reason, when the value of the temporary variable WMs is originally 0.5 or less, it is forcibly fixed to a value of 0.5 (steps S41 and S42).
[0053]
This is because increasing the amplitude of the watermark as much as possible within the range where the distortion of the image cannot be grasped visually is beneficial in increasing the strength of the watermark. If the value of the temporary variable WMs is 0.5, the final output amplitude WMs is
WMs = 0.5 x TB (Equation 9)
Is equivalent to
[0054]
In general,
0 <L ≤ 0.5 (Equation 10)
Is set to a constant L, and the value of the temporary variable WMs calculated by Equation 7 is
WMs <L (Equation 11)
May be fixed to the value of the constant L forcibly. If the value of the temporary variable WMs is L, the final output amplitude WMs is
WMs = L × TB ・ ・ ・ ・ (Formula 12)
Is equivalent to Among the values of various constants L (≦ 0.5), L = 0.5 is the most desirable form in that the strength of the watermark is maximized.
[0055]
More generally, when the condition of Equation 11 is satisfied, the temporary variable WMs is
L ≦ K ≦ 0.5 ・ ・ ・ ・ (Formula 13)
May be set to a constant K given by. If the value of the temporary variable WMs is K, the final output amplitude WMs is
WMs = K × TB ・ ・ ・ ・ (Formula 14)
Is equivalent to
[0056]
When the process of step S44 ends, the calculated amplitude WMs is output, and the process of step S26 is completed. Thereafter, the process proceeds to step S27 in FIG.
[0057]
In step S27, the binary data constituting the watermark WM is converted from a format expressed by "0" and "1" to a format expressed by "-1" and "1". Furthermore, the watermark WM is duplicated for each identification number IND. Next, in step S28, the watermark WM is multiplied by a random number PN. In subsequent step S29, the watermark WM is multiplied by the amplitude WMs calculated in step S26. As a result, a final watermark WM to be embedded is obtained.
[0058]
Next, the process of step S30 is repeatedly executed for each identification number IND. In step S30, the watermark WM is superimposed on the luminance component Ysp for each identification number IND, and the luminance component Ysp ′ after watermark embedding is obtained. When the watermark is embedded for all the identification numbers IND, step S4 ends, and the process proceeds to step S5 (FIG. 1). In addition, since the process of step S24, S27-S30 is well-known in the conventional watermark encoding method (FIG. 10), detailed description about these is abbreviate | omitted.
[0059]
【The invention's effect】
In the method of the first invention, since the original image data is stored in the form of a compressed image with a small data amount, the original image to be stored in a distribution source that normally needs to store many types of original image data. The amount of data is saved. That is, the burden on the distribution source for storing the original image data is reduced. Moreover, in order to search for spectral components that can be embedded in the watermark, random fluctuations are added to the spectral data, and then quantization is performed in a larger quantization step than steps (a) and (e). Spectral components in which the quantization coefficient value obtained thereby is not zero are searched for.
[0060]
For this reason, the deterioration of the image quality due to the deviation appearing between the two spectral data obtained in the steps (a) and (c), which is an irreversible process, is suppressed. In addition, since the watermark is embedded in the spectrum component before the fluctuation is superimposed, the influence of the fluctuation does not remain in the distributed image data.
[0061]
In the method of the second invention, since the original image data is stored in the form of a compressed image with a small data amount, the original image to be stored in a distribution source that normally needs to store a large number of types of original image data. The amount of data is saved. That is, the burden on the distribution source for storing the original image data is reduced. In addition, since the amplitude of the watermark to be superimposed is set to a value larger than ½ times the quantization step used in the quantization of steps (a) and (e), the embedded watermark is quantized. The inconvenience of disappearing after conversion can be avoided. As a result, the problem of a decrease in the decoding accuracy rate is also avoided.
[0062]
In the method of the third invention, the amplitude of the watermark to be overlaid is set so as not to fall below a constant L (≦ 1/2) times the value of the quantization table. This increases the strength of the watermark within a range where the distortion of the image is not recognized visually.
[0063]
In the method of the fourth invention, it can be determined whether or not the collected image data is an illegal copy of the image data distributed after being encoded by the method of the first to third inventions. .
[Brief description of the drawings]
FIG. 1 is a flowchart of a watermark encoding method according to an embodiment.
FIG. 2 is a flowchart of a watermark decoding method according to the embodiment.
FIG. 3 is a timing chart of the procedure of FIGS. 1 and 2;
FIG. 4 is a graph for explaining a first factor of a decrease in the decoding accuracy rate.
FIG. 5 is a graph for explaining a second factor of a decrease in the decoding accuracy rate.
6 is a flowchart showing an internal flow of step S4 in FIG.
FIG. 7 is a graph for explaining the reason why the first factor is eliminated.
FIG. 8 is a graph showing an internal flow of step S23 of FIG.
FIG. 9 is a graph showing an internal flow of step S26 of FIG.
FIG. 10 is a flowchart of a conventional watermark encoding method.
FIG. 11 is a flowchart of a conventional watermark decoding method.
12 is a timing chart of the procedures of FIGS. 10 and 11. FIG.
[Explanation of symbols]
δ Fluctuation
Δ Quantization step
Q Quantization coefficient
SPC, SPC 'spectrum data
SC scaling factor
TB quantization table
WM watermark

Claims (4)

In a watermark encoding method for embedding a watermark in original image data,
(a) converting the original image data from an uncompressed original image format into a compressed image data format through image compression including conversion to spectral data and quantization;
(b) storing the original image data in the compressed image data format;
(c) converting the original image data from the format of the compressed image data to spectral data through image expansion including inverse quantization as an inverse operation of the quantization;
(d) superimposing a watermark on the spectral data obtained in step (c);
(e) converting the spectral data with the watermark superimposed in the step (d) into a format of the compressed image data through image compression including the quantization;
(f) delivering the compressed image data obtained by converting in the step (e),
Step (d)
(d-1) superimposing random fluctuations of the code on the spectrum data obtained in the step (c);
(d-2) Quantize the spectral data on which random fluctuations are superimposed in the step (d-1), using a scaling coefficient that is larger than the scaling coefficient used in the quantization. To obtain a quantization coefficient,
(d-3) configuring the spectral data, and selecting a watermark embedding target from among spectral components that give a non-zero value as the quantization coefficient;
(d-4) superimposing the watermark on a value before the fluctuation of the spectral component selected in the step (d-3) is superimposed;
A watermark encoding method comprising: In a watermark encoding method for embedding a watermark in original image data,
(a) converting the original image data from an uncompressed original image format into a compressed image data format through image compression including conversion to spectral data and quantization;
(b) storing the original image data in the compressed image data format;
(c) converting the original image data from the format of the compressed image data to spectral data through image expansion including inverse quantization as an inverse operation of the quantization;
(d) superimposing a watermark on the spectral data obtained in step (c);
(e) converting the spectral data with the watermark superimposed in the step (d) into a format of the compressed image data through image compression including the quantization;
(f) delivering the compressed image data obtained by converting in the step (e),
Step (d)
(d-1) A watermark encoding method comprising: setting the amplitude of the watermark to be superimposed to a value larger than ½ times the quantization step used in the quantization. The watermark encoding method according to claim 2,
Step (d)
(d-2) When the amplitude set in the step (d-1) is smaller than L times the value of the quantization table with respect to a predetermined constant L (≦ 1/2) set in advance And a step of setting the amplitude to a value that is a predetermined constant K (L ≦ K ≦ 1/2) times the value of the quantization table. In the watermark decoding method for extracting the watermark from the collected image data,
(A) By executing the same step as the step (c) on the original image data stored by the step (b) of the watermark encoding method according to any one of claims 1 to 3, Obtaining spectral data;
(B) converting the collected image data into spectral data;
(C) calculating the difference between the two spectral data obtained in the steps (A) and (B), and extracting the watermark from the spectral data obtained in the step (B); A watermark decoding method comprising:

Images