JP3985903B2 - Image compression method, digital watermark embedding method using the same, and program - Google Patents

Description

  The present invention relates to digital watermark embedding, and more particularly to an image compression technique for embedding a digital watermark.

  There is a digital watermark embedding method for embedding information by applying minute luminance modulation to pixels of an image (for example, see Non-Patent Document 1). In this case, in the conventional technique, a digital image in which a digital watermark is embedded is generated and watermark information is read from this digital image. Analog output such as printing or screen display of the captured digital image, capturing the digital image with digital watermark by capturing it with a camera, etc., and reading the watermark information from the captured image. Even in this case, embedding of the digital watermark itself was performed by the same method.

Kokoo Matsui, 1998, Morikita Publishing, ISBN4-627-82511-X "Basics of Digital Watermarking", "3.1-pixel replacement type digital watermarking" p38-p41

  When a digital image is read using a camera or the like from a display device that displays a watermarked printed matter or watermark image, and the digital watermark is read in the prior art, a minute amount due to watermark modulation expressed in the digital image data immediately after the digital watermark is embedded. In some cases, the luminance change attenuates and blurs in the printing process, the display process, or the camera photographing process, and the watermark reading fails.

  The present invention provides an image compression method and an electronic watermark embedding method for embedding an electronic watermark, in which the electronic watermark can be correctly read out from the captured image without being affected by the blur even when analog data conversion is performed in the middle. As well as its program.

  When an image is expressed by digital data, it is generally expressed by RGB values for each pixel (for example, 0 ≦ R, G, B ≦ 255). In addition, expressions in YCbCr, YUV, and XYZ obtained by converting RGB with a prescribed conversion formula are also used. These are reversible transformations. When embedding the digital watermark by luminance modulation, for example, the image data is converted to YUV, and the luminance component Y is modulated by the digital watermark to obtain a digital watermark embedded image. If necessary, it is then converted into an expression format such as RGB.

  When the watermark-embedded image data is output by an output device such as a printing device or a display, the image data is converted in accordance with a color space unique to the output device, and colors that cannot be expressed are rounded to the nearest color in the color space. Further, when an image of the output device is captured by an input device such as a camera or a scanner, conversion into a color space unique to the input device is performed here, and rounding of an unrepresentable color occurs. As a result, the image data obtained by capturing the watermark-embedded image output by the output device with the input device is converted compared to immediately after embedding, and the degree of data conversion particularly in the peripheral portion of the color space is large. Hereinafter, conversion of image data in a system including an output device and an input device is referred to as analog conversion.

  The method of embedding a digital watermark by luminance modulation is not significantly affected by the change in color due to analog conversion, but is greatly affected by the change in luminance. In particular, the black and white parts of the image, that is, the part corresponding to the peripheral part in the color space, are easily affected. When a digital watermark is embedded in this part, the brightness that is expressed on the digital image data immediately after the embedding is small. The change decreases as it passes through the output device and the input device. When the watermark is read, the change disappears, and the watermark may not be read. This is a phenomenon called “collapse” of so-called black or white.

  In order to prevent digital watermarks from being affected by “crushing”, a method that does not embed a watermark in an image that is likely to be crushed and a method that does not embed the watermark before embedding the watermark and analog conversion are used. A way to prevent it is possible. However, the former is not preferable because the area in which the watermark can be embedded differs depending on the image, and depending on the image, the watermark cannot be embedded at all. Therefore, the latter method is used in the present invention.

(1) Determining the degree of luminance compression In order to determine the degree of luminance compression, it is necessary to know in advance the range of luminance that is lost by analog conversion, that is, where “crushing” occurs. Since analog conversion is a property unique to a device, it is necessary to measure the characteristics of each device to be used. In the present invention, the reference color chart shown in FIG. 1 is prepared. In FIG. 1, an annular band 11 is composed of data obtained by gradually changing the minimum value to the maximum value of luminance that can be expressed by digital data in the circumferential direction. The reason for the annular shape is that when the color chart 10 is photographed with a camera, it is not affected by the cosine fourth law, which appears brighter in the center of the photograph and darker in the periphery. The background 12 of the toroidal band is square and halftone gray. The reason for the square is to facilitate reverse projection conversion after shooting, and the reason for the gray to be halftone is to minimize the influence of the auto white balance of the camera.

  The color chart 10 is printed by a printing device that is a target output device or displayed on a display. Next, the printed or displayed color chart image is captured by a camera or a scanner as a target input device. When the image is taken by the camera, the image has undergone projective transformation, so reverse projection transformation is performed using the four corners of the outer square as a clue to restore the same shape as the original data (color chart data) of the color chart 10. In the case of the scanner, affine transformation has been performed, so reverse affine transformation is similarly applied to the four corners of the square to restore the original data shape. Next, the luminance of each pixel of the annular band of the color chart 10 is acquired from the capture data and the original data, and the minimum and maximum values of the luminance of the annular band are obtained. Both the output device and the input device The minimum luminance l and the maximum luminance h that can be expressed in the original data are obtained.

  FIG. 2 shows an example in which the brightness of the capture data is plotted on the vertical axis and the brightness of the original data is plotted on the horizontal axis. In FIG. 2, it can be seen that the luminance range of 1 or less and the luminance range of h or more of the original data cannot be expressed by the capture data, and the image in this range is “crushed”. Although a graph is used for the explanation here, it is not necessary to create a graph to actually obtain l and h, and the minimum and maximum luminance values of the annular band of the capture data are obtained and the corresponding original data You only need to get the value.

  l and h mean the minimum luminance and the maximum luminance that can be expressed by the compressed image, respectively, and if the image is compressed so that the luminance of the compressed image falls between l and h, “crushing” does not occur. Assuming that the maximum intensity modulation of the digital watermark is ± m, when embedding a digital watermark, the luminance of each pixel may vary by up to ± m compared to before embedding. The minimum value of the image luminance is L = 1 + m, and the maximum value is H = hm.

(2) Luminance compression function The compression function f (x) for compressing the original image so as to be in the luminance range L to H has many variations, and it is a necessary condition to satisfy the following expression (1).

ただし、ｘ_ｍｉｎは元画像の輝度最小値
ｘ_ｍａｘは元画像の輝度最大値
ｆ（ｘ）の条件を図示すると図３のようになり、Ｌ点とＨ点を通り微分値が負にならない関数であることが条件である。微分値が負になるのを禁じているのは、元画像と圧縮後画像で２点間の輝度の大小が逆転する場合を除外するためである。

Where x _min is the minimum luminance of the original image
FIG. 3 shows the condition of the maximum luminance value f (x) of the original image , where x _max is a function that does not have a differential value passing through the L point and the H point. The reason why the differential value is prohibited from being negative is to exclude the case where the magnitude of the luminance between two points is reversed between the original image and the compressed image.

(3) Evaluation of luminance compression function The luminance compression function f (x) that satisfies the condition of the above expression (1) has variations, but is the same in that the original image is compressed from the luminance range L to H. In terms of whether the watermark is stored without being crushed, the performance is the same regardless of which function is used. However, in terms of image quality, performance varies greatly depending on how f (x) is selected. In luminance compression, since some modification is made to the original image, it is difficult for the quality to not deteriorate at all, but it is desirable to minimize the quality deterioration. The following three evaluation criteria are used in order to select the one with small quality degradation from the variations of f (x).

(A) Simple luminance difference It is assumed that the luminance difference between two pixels in the same position of the original image and the compressed image is obtained, and the smaller the sum of the luminance differences in the entire image, the less the quality deterioration due to compression.

(B) Uniform color space color difference A perceptual difference between two pixels in the same position in the original image and the compressed image is called a color difference, and it is assumed that the smaller the sum of the color differences in the entire image, the less the deterioration in quality due to compression. As a method for quantitatively expressing the color difference, a method using the Euclidean distance between two pixels plotted in the CIELAB color space or CIELV color space recommended by the International Commission on Illumination (CIE) is generally used. However, since the simplicity of the mathematical formula is obtained for both the CIELAB color space and the CIELV color space, it is insufficient to accurately represent the color difference. Therefore, the difference between the original image and the compressed image is only the brightness, and the CIELAB color space and the CIELV color space are both 1/3 power laws (the visual intensity perceived by humans is proportional to 1/3 power of visual stimuli). In consideration of the point that is experimentally obtained based on the above, the difference between the luminances of the two points raised to the 1/3 power is used as the color difference.

(C) Local compression rate The ratio Δx _c / Δx (referred to as the local compression rate) between the luminance difference Δx between any two points in the original image and the luminance difference Δx _c between the two corresponding points in the compressed image is called zero. If they are close to each other, the light and shade that existed in the original image is not preferable because the image after compression attenuates or disappears. This is also a kind of “collapse”. As long as compression is performed, it is inevitable that the local compression rate becomes a small value. However, even if a part of the image is controlled so as not to approach zero, it is necessary to prevent “collapse”. Since the luminance range after compression is suppressed by L and H, in order to prevent the minimum value of the local compression rate from approaching zero, it is only necessary to reduce the fluctuation in the image of the local compression rate. Therefore, the standard deviation representing the fluctuation of the local compression rate is used as the evaluation criterion.

(4) Determination of Optimal Luminance Compression Function The luminance compression function f (x) with the best quality is determined using the above three evaluation criteria and combinations thereof. In the embodiment , one of the following two methods is used.

The first method is to prepare a satisfy the condition of previous equation (1) some of the compression function f (x) in advance, the best quality using the combination with three evaluation criteria of the f ( x) is selected.

  The second method uses the following equation (2) as a compression function, and determines the value of a that gives the best quality using the above three evaluation criteria and combinations thereof.

ここで、上記式（２）で表わされるｆ（ｘ）は任意のａにおいて先の式（１）を満たすので、輝度圧縮関数としての働きはａの値によらず同等である。したがって、画像の品質劣化が最小になるようにａの値を自由に選ぶことができる。式（２）の圧縮関数は調整変数がａのみなので、調整が容易である特徴がある。上記の３つの評価基準とその組み合わせを用いて品質が最良になるようａを選択すればよい。なお、いずれの評価基準を用いる場合も、元画像の絵柄によって最適なａは微妙に異なるため、任意画像に対してほぼ十分な画像品質を与えるａをひとつ選んでも良いし、画像毎に最適なａを選んでもよい。

Here, since f (x) represented by the above formula (2) satisfies the above formula (1) at an arbitrary a, the function as a luminance compression function is the same regardless of the value of a. Therefore, the value of a can be freely selected so that image quality degradation is minimized. The compression function of Expression (2) has a feature that adjustment is easy because the adjustment variable is only a. A may be selected so that the quality is the best using the above three evaluation criteria and combinations thereof. Note that, regardless of which evaluation criterion is used, the optimum a differs slightly depending on the pattern of the original image. Therefore, one a that gives almost sufficient image quality to an arbitrary image may be selected, or the optimum for each image. You may choose a.

(5) Digital watermark embedding The luminance of the original image is compressed with the compression function f (x) determined as described above, and the digital watermark is embedded in the compressed image using luminance modulation. Here, it is assumed that the luminance value of each pixel of the image after embedding varies by a maximum of ± m compared to before embedding by the digital watermark of the luminance modulation method. In addition, it is assumed that the luminance range that can be reproduced by analog conversion of the device to be used is 1 to h in advance. Then, assuming that L = 1 + m and H = hm, the luminance range of the original image is suppressed to L to H by luminance compression before embedding the watermark, so that the image after embedding the watermark does not exceed the luminance range 1 to h. Therefore, it is possible to suppress the “collapse” caused by analog conversion from adversely affecting the watermark.

  In the present invention, when embedding a digital watermark into an image, the image is brightness-compressed in advance so as not to cause blurring, and the watermark information is embedded in the compressed image. When the captured image is captured with an analog input device, no blur occurs even if the high-luminance or small-luminance portion disappears due to the characteristics of each device. It can be read out.

  FIG. 4 shows the overall processing flow of the image compression method and digital watermark embedding method for embedding a digital watermark according to the present invention. First, a minimum expressible luminance l and a maximum expressible luminance h when both the output device and the input device are passed using a color chart are obtained (S110). Next, assuming that the digital watermark luminance modulation maximum intensity is ± m, the minimum luminance L and the maximum luminance H of luminance compression are determined as L = 1 + m, H = hm (S120), and the compression with the best image quality evaluation value is performed. A function (luminance compression function) f (x) is obtained (S130). Next, using the compression function f (x) thus obtained, the original image is subjected to luminance compression so that the compressed luminance distribution is suppressed between L and H (S140). Then, an electronic watermark is embedded in the compressed image using luminance modulation (S150). Below, the Example is described about the process of step S110, S130 which is the characteristics of this invention.

<Analog conversion characteristics measurement>
FIG. 5 shows a flowchart of an embodiment for obtaining the expressible minimum luminance l and the expressible maximum luminance h in the analog conversion of the arbitrary output device and the arbitrary input device. First, the color chart of FIG. 1 is output by the target output device (S201). Next, the output color chart image is captured by a camera or scanner which is a target input device, and is input to an image compression processing or digital watermark embedding processing device (personal computer, etc.) (S202). Step S203 and subsequent steps are processing within the processing apparatus. If the image is taken with a camera, the image has undergone projective transformation, so reverse projection transformation is performed using the four corners of the square in the outer gray part of the color chart image to restore the same shape as the original data of the color chart ( S203, S204, S205). In the case of a scanner instead of a camera, affine transformation has been performed, so reverse affine transformation is similarly performed using the four corners of the square as a clue to return to the original data shape (S203, S204, S206). When capture data having the same shape as the original data shape is obtained, the pixel luminance at the center position of the band width of the annular band is acquired from the capture data in the circumferential direction, and the minimum value and maximum value thereof are obtained (S207). Then, the minimum representable luminance l and the maximum representable luminance h obtained through both the output device and the input device are obtained (S208). That is, as shown in FIG. 2, the original data value at the position corresponding to the minimum value position is l, and the original data value at the position corresponding to the maximum value position is h.

<Optimal compression function calculation>
Using the above three evaluation criteria (simple luminance difference, uniform space color difference, local compression rate), or a combination thereof, the compression function f (x) with the best quality is obtained. Several examples are shown below.

  FIG. 6 shows a compression function f (x) having the best image quality evaluation value using a simple luminance difference as an evaluation criterion from among variations of an arbitrary compression function f (x) that satisfies the condition of the expression (1). It is a processing flowchart which shows the Example which determines. First, one compression function f (x) is selected from several variations of an arbitrary compression function f (x) that satisfies the condition of Expression (1) (S301), and the selected compression function f (x) In step S302, the original image is compressed to generate a post-compression image. Next, the luminance difference between the same position pixels (pixels) of the original image and the compressed image is obtained for the entire image (all pixels) and integrated (S303). That is, the sum of luminance differences between the original image and the compressed image is calculated. Then, the sum (integral value) of the luminance differences is compared with a threshold value determined in advance by statistics or the like (S304), and if the integral value <threshold value, the selected compression function f (x) is determined as the optimum compression function. (S305). On the other hand, if the integral value ≧ the threshold value, another compression function f (x) is selected (S306), and the processes after step S302 are repeated.

  Instead of the threshold determination, the compression function f (x) may be sequentially selected and the integration value thereof may be held, and the compression function corresponding to the minimum integration value may be finally selected from the selection. . In addition, if the integration value of any compression function f (x) is equal to or greater than the threshold value, the processing may be given up or may be retried in the manner of another embodiment.

  In FIG. 7, the compression function f (x) of the above equation (2) is used as the compression function f (x), a is variable, and the simple luminance difference is used similarly to the compression function f ( It is a process flowchart which shows the Example which determines x). The compression function f (x) is defined as f (x) in the above equation (2) (S401). First, the value of a is initially set to, for example, a = 0.1 (S402), and f ( The original image is compressed using x) to generate a compressed image (S403). Next, the luminance difference between the same position pixels (pixels) of the original image and the compressed image is obtained for the entire image (all pixels) and integrated (S404). Then, the integral value is compared with a predetermined threshold value (S405). If the integral value <threshold value, the value of a at that time is determined to be optimum, and f (x) using the value a is optimum compression function. Is determined (S406). On the other hand, when the integral value ≧ the threshold value, the value of a in f (x) is changed to, for example, a = 0.2 (S407), and the processes after step S403 are repeated.

  Also in this embodiment, instead of threshold determination, each integral value obtained by sequentially changing a (for example, a = 0.1 to 2) is held, and the minimum integral value is supported from among them. The value of a to be selected may be selected, and f (x) in Expression (2) obtained by substituting the value of a may be determined as the optimum compression function. In addition, when the integral value is greater than or equal to the threshold value for any value of a, the processing may be given up or may be retried in the manner of another embodiment.

  When the uniform color space color difference is used as the evaluation criterion, the luminance of each identical pixel in the original image and the compressed image is set to 1 as the luminance difference between the original image and the compressed image in step S303 in FIG. 6 or step S304 in FIG. A difference between the values raised to the third power is obtained and integrated for all pixels, and the processing flowchart is omitted.

  FIG. 8 shows that the image quality evaluation value is the best using the standard deviation representing the fluctuation of the local compression rate as the evaluation criterion from among the variations of the arbitrary compression function f (x) satisfying the condition of the expression (1). It is a processing flowchart which shows the Example which determines the compression function f (x).

First, for each pixel of the original image, a luminance difference Δx between the pixel and all other pixels is calculated (S501). If the number of pixels in the image is x, Δx can be x (x−1). Next, one compression function f (x) is selected from several variations of the compression function f (x) that satisfies the condition of Expression (1) (S502), and the selected compression function f (x) The original image is compressed to generate a compressed image (S503). Then, for this compressed image, as in the case of the original image, a luminance difference Δx _c between the pixel and all other pixels is calculated for each pixel (S504). Again, Δx _c can be x (x−1). Next, calculate the ratio [Delta] x c _/ [Delta] x of the corresponding [Delta] x and [Delta] x _c of the compressed image after the original image, the standard deviation value indicating the variation (S505). Then, the threshold value of the standard deviation value is determined (S506). If the standard deviation value <the threshold value, the selected compression function f (x) is determined as the optimum compression function (S507). On the other hand, when the standard deviation value ≧ the threshold value, another compression function f (x) is selected (S508), and the processes after step S503 are repeated.

  Also in the present embodiment, instead of threshold determination, the standard deviation value of each compression function f (x) may be held, and the compression function corresponding to the minimum standard deviation value may be finally selected from among them. . In addition, when the standard deviation value of any compression function f (x) is greater than or equal to the threshold value, the processing may be abandoned or may be retried in the manner of another embodiment.

  FIG. 9 uses f (x) of the above equation (2) as the compression function f (x), a is variable, and uses the standard deviation value representing the fluctuation of the local compression rate as the evaluation value criterion as described above. 10 is a flowchart illustrating an example in which a compression function f (x) having an optimum image quality evaluation value is determined.

First, for each pixel of the original image, a luminance difference Δx between the pixel and all other pixels is calculated (S601). Next, the compression function f (x) is defined as f (x) in the previous equation (2) (S602), the value of a is initialized (for example, a = 0.1) (S603), and the original image is set. Are compressed, and a compressed image is generated (S604). Then, for this compressed image, as in the case of the original image, a luminance difference Δx _c between the pixel and all other pixels is calculated for each pixel (S605). Next, calculate the ratio [Delta] x / [Delta] x _c of the corresponding [Delta] x and [Delta] x _c of the compressed image after the original image, the standard deviation value indicating the variation (S606). Then, the standard deviation value is determined as a threshold value (S607). If standard deviation value <threshold value, the value of a at that time is set as the optimum value, and f (x) is determined as the optimum compression function (S608). On the other hand, if the standard deviation value ≧ the threshold value, the value of a in f (x) is changed (for example, a = 0.2) (S609), and the processes after step S604 are repeated.

  Also in this embodiment, instead of the threshold value determination, the standard deviation value when the value of a is sequentially changed is held, and the value of a corresponding to the minimum standard deviation value is selected from the standard deviation values. The compression function f (x) in equation (2) obtained by substituting the value of may be an optimal compression function. If the standard deviation value is greater than or equal to the threshold value for any value of a, the processing may be abandoned, or another attempt may be made in the manner of another embodiment.

  Although the processing flowchart is omitted, the compression function f (x) that provides the best image quality evaluation value using a combination of the simple luminance difference and the local compression rate may be obtained as follows. For example, when selecting the optimum f (x) from any variation of f (x) that satisfies the condition of Equation (1), the integral value obtained in S303 of FIG. The standard deviation value obtained in S505 is set as D, aE + bD is calculated (a and b are arbitrary coefficients), and f (x) that minimizes the value is selected. When f (x) in equation (2) is used to determine the optimum f (x) with a being variable, similarly, the integral value obtained in S404 of FIG. The standard deviation value obtained in S606 is calculated as D, aE + bD is calculated, a is determined so that the value is minimized, and the optimum f (x) is determined.

Next, a specific example of the compression function is shown.
[Example 1]
FIG. 10 shows values before and after compression using f (x) = (HL) x + L, (0 ≦ x ≦ 1) as the compression function. Since this function is linear, it is easy to calculate. However, as shown in FIG. 11, since the luminance difference between the original image is large in the small range and the large range of x, the color difference in the uniform color space in the small range of x is different depending on the third power that is human visual characteristics. Has the disadvantage of becoming larger.

[Example 2]
In FIG. 12, as a compression function, f (x) = L, (0 ≦ x ≦ 1)
f (x) = x, (L ≦ x ≦ H)
f (x) = H, (H ≦ x ≦ L)
The values before and after compression using are shown. This function has the smallest uniform color space color difference. However, in the range where x is L or less and the range is H or more, there is a problem that the shading expressed in the original image is “crushed”.

[Example 3]
FIG. 13 shows values before and after compression using f (x) in Expression (2) as the compression function. This function shows various properties depending on a. FIG. 14 is a plot of the integral value of the entire image of the uniform color space color difference with a being changed for an image. In this image, the vicinity of a = 1.1 minimizes the color difference integration. FIG. 15 is a plot of local compression rates for the same image. The fluctuation of the local compression rate is small near a = 0.5. In order to make the uniform color space color difference and the local compression ratio compatible, the linear combination 0.5E + 0.5D of the integral value E for the entire image of the uniform color space color difference and the standard deviation value D for the entire image of the local compression ratio is Choose the a that minimizes. The coefficient of the linear combination is changed depending on whether the color difference or the local compression rate is important.

<System example>
FIG. 16 shows a specific system example of an embodiment of an image compression method and a digital watermark embedding method according to the present invention. In this example, a printer is used as an output device, and a camera with a built-in mobile phone is used as an input device. The color chart shown in FIG. 1 is printed by the printer 702, and the print result 703 is photographed by the camera 704 built in the mobile phone. This photographed image (color chart image) and original color chart data 701 are input to a personal computer (PC). In the PC, the minimum luminance position and the maximum luminance position of the annular band of the photographed image are obtained, and the expressible minimum luminance l and maximum luminance h are obtained from the corresponding original data position (705). If the mobile phone has a calculation function, l and h may be obtained directly from the mobile phone instead of the PC. Next, in the PC, the target minimum luminance L and the maximum luminance H for luminance compression are determined by L = 1 + m and H = hm from the known luminance modulation maximum intensity ± m of the luminance modulation digital watermark (706, 707). Next, in the present embodiment, the compression function is defined as f (x) in the previous equation (2), and luminance compression is performed on the original image 710 by changing, for example, from a = 0.1 to 2.0. (708). Then, for example, a linear combination 0.5E + 0.5D of an integral value E with respect to the entire image of the uniform color space color difference and a standard deviation value D with respect to the entire image of the local compression rate is obtained from the compressed image, and this is minimized. The value of a is determined (709). Next, the original image 710 is again subjected to luminance compression (711) using f (x) of the expression (2) into which the obtained a is substituted, and a digital watermark is embedded in the obtained image (712). As a result, a digital watermark embedded image 713 having a luminance range of L to H is obtained. Even if this image is printed by a printer and the print result is taken by a camera with a built-in mobile phone, no blur occurs and the digital watermark can be read out appropriately. Needless to say, the optimum f (x) may be determined by applying the embodiments shown in FIGS.

  Needless to say, the processing procedures shown in FIGS. 4 to 9 can be configured by a computer program, and the program can be executed by the computer. In addition, a program for causing a computer to execute the processing procedure may be recorded and stored on a recording medium readable by the computer, for example, FD, MO, ROM, memory card, CD, DVD, removable disk, etc. Can be provided, and the program can be distributed through a network such as the Internet.

The figure which shows the color chart used for this invention. The figure which shows the luminance conversion by analog conversion. The figure explaining the conditions of compression function f (x). 3 is an overall processing flowchart of image compression and digital watermark embedding according to the present invention. The processing flowchart of one Example of an analog conversion characteristic measurement. The processing flowchart of the 1st Example which calculates | requires the optimal compression function f (x). The processing flowchart of the 2nd Example which calculates | requires the optimal compression function f (x). The process flowchart of the 3rd Example which calculates | requires the optimal compression function f (x). The processing flowchart of a 4th Example which calculates | requires the optimal compression function f (x). The graph which shows a linear compression function. The figure which shows the uniform color space color difference of a linear compression function. The graph which shows the uniform color space minimum color difference function. The figure which shows the example of non-linear compression. The figure which shows the change of a uniform color space color difference integrated value. The graph which shows a local compression rate. 1 is a diagram showing an embodiment of a digital watermark embedding system according to the present invention.

Explanation of symbols

10 Color chart S110 Analog conversion characteristic measurement S120 Determination of minimum / maximum brightness S130 Calculation of optimal compression function S140 Original image compression S150 Digital watermark embedding

Claims (7)

An image compression method for embedding a digital watermark,
The output device outputs a color chart with a circular band in which the luminance gradually changes in the circumferential direction from the minimum value that can be digitally expressed to the maximum value , and the background is a gray of gray. Capturing the color chart image with the input device and converting it into digital data, obtaining the minimum and maximum luminance values on the digital data, and expressing it through both the output device and the input device Determining a minimum luminance l and a maximum luminance h;
Obtaining a minimum value L = l + m and a maximum value H = hm of the image luminance after compression in consideration of embedding of the digital watermark, with the luminance modulation maximum intensity of the digital watermark taken as ± m;
The following conditions

Where x _min is the minimum luminance of the original image
x _max is the maximum luminance of the original image
A compression function f (x) that satisfies

Then, an image obtained by compressing the brightness of the original image by sequentially changing the value of a within a predetermined range is obtained, and the integral value for the entire image of the brightness difference between two pixels at the same position in the original image and the compressed image is minimized. selecting a value of a;
A step of obtaining a compressed image in which the luminance distribution is pressed between L and H by performing luminance compression on the original image using the compression function f (x) of the equation (2) into which the value of the selected a is substituted. When,
An image compression method characterized by comprising: The image compression method according to claim 1,
An image compression method characterized by using a difference between values obtained by raising the luminance of the original image and the compressed image to the 1/3 power as the luminance difference between the original image and the compressed image. An image compression method for embedding a digital watermark,
The output device outputs a color chart with a circular band in which the luminance gradually changes in the circumferential direction from the minimum value that can be digitally expressed to the maximum value, and the background is a gray of gray. Capturing the color chart image with the input device and converting it into digital data, obtaining the minimum and maximum luminance values on the digital data, and expressing it through both the output device and the input device Determining a minimum luminance l and a maximum luminance h;
Obtaining a minimum value L = l + m and a maximum value H = hm of the image luminance after compression in consideration of embedding of the digital watermark, with the luminance modulation maximum intensity of the digital watermark taken as ± m;
The following conditions

Where x _min Is the minimum brightness of the original image
x _max Is the maximum brightness of the original image
A compression function f (x) that satisfies

To obtain an image in which the luminance of the original image is compressed by sequentially changing the value of a within a predetermined range, and the luminance difference Δx between any two points of the original image and the luminance difference between two corresponding points of the compressed image Δx _c Ratio Δx _c Selecting a value of a that minimizes the standard deviation value representing the variation in the image of / Δx;
A step of obtaining a compressed image in which the luminance distribution is pressed between L and H by performing luminance compression on the original image using the compression function f (x) of the equation (2) into which the value of the selected a is substituted. When,
An image compression method characterized by comprising: An image compression method for embedding a digital watermark,
The output device outputs a color chart with a circular band in which the luminance gradually changes in the circumferential direction from the minimum value that can be digitally expressed to the maximum value, and the background is a gray of gray. Capturing the color chart image with the input device and converting it into digital data, obtaining the minimum and maximum luminance values on the digital data, and expressing it through both the output device and the input device Determining a minimum luminance l and a maximum luminance h;
Obtaining a minimum value L = l + m and a maximum value H = hm of the image luminance after compression in consideration of embedding of the digital watermark, with the luminance modulation maximum intensity of the digital watermark taken as ± m;
The following conditions

Where x _min Is the minimum brightness of the original image
x _max Is the maximum brightness of the original image
A compression function f (x) that satisfies

And an integral value E for the entire image of the luminance difference e between the two pixels at the same position in the original image and the compressed image is obtained by sequentially changing the value of a within a predetermined range and compressing the luminance of the original image. The luminance difference Δx between any two points in the original image and the luminance difference Δx between two corresponding points in the compressed image _c Ratio Δx _c Selecting the value of a that minimizes the linear combination of arbitrary coefficients with the standard deviation value D representing the variation of / Δx;
A step of obtaining a compressed image in which the luminance distribution is pressed between L and H by performing luminance compression on the original image using the compression function f (x) of the equation (2) into which the value of the selected a is substituted. When,
An image compression method characterized by comprising: In a digital watermark embedding method for embedding information by applying minute luminance modulation to pixels of an image,
5. An electronic watermark embedding comprising: compressing an original image using the image compression method according to claim 1; and embedding an electronic watermark using luminance modulation in the compressed image. Method. The program for making a computer perform the image compression method of any one of Claims 1-4. A program for causing a computer to execute the digital watermark embedding method according to claim 5.

Images