A method for producing and extracting digital blind watermark

Abstract

The present invention discloses a method for making and extracting a digital blind watermark, which comprises the following steps: S1, obtaining a monochrome area in a picture, if the picture contains a pure monochrome area, directly recording the monochrome area; if the picture contains only an approximate monochrome area, fine-tuning the color of the approximate monochrome area to produce a monochrome area; S2, dividing the found monochrome area into N rows and M columns of squares, each square pixel is a*a, and the channel color value of the monochrome area is 2 or +2, 2 represents binary 0, +2 represents binary 1, to form multiple binary hidden information. The present invention is simple and clear, and at the same time, it resists the damage of hidden information caused by partial post-processing of the picture to a certain extent.

Description

Technical Field

[0001] The invention relates to a method for producing and extracting a digital blind watermark. Background Art

[0002] Currently, there are many ways to implement digital blind watermarking on the market, such as frequency domain transformation method, glyph transformation method, color LSB transformation method, Patchwork method, neural network-based algorithm, etc.

[0003] Among the various digital blind watermarking technologies currently available, the frequency domain transformation method and the neural network-based algorithm are relatively more robust. After processing the image, such as blurring, scaling, adjusting the image brightness and contrast, etc., they can restore the hidden information carried by the image to a certain extent. However, the above two methods are relatively complex to implement.

[0004] The color LSB transformation method is simple and intuitive to implement, but the robustness of this blind watermarking technology is too poor. Almost any modification to the image, such as scaling, adjusting the image brightness, blurring, etc., will make the hidden information in the original image irreversible. Summary of the invention

[0005] In order to solve the above problems, the present invention provides a method for producing and extracting a digital blind watermark which is simple and clear and can resist to a certain extent the damage of hidden information caused by partial post-processing of the image.

[0006] A method for producing a digital blind watermark comprises the following steps:

[0007] S1. Obtain the monochrome area in the image. If the image contains a pure monochrome area, directly record the monochrome area; if the image only contains an approximate monochrome area, fine-tune the color of the approximate monochrome area to produce a monochrome area. The color fine-tuning method is as follows: the first step is to determine the starting position of the monochrome area. The color difference between the current pixel a1 and the subsequent pixel a2 does not exceed β; the color values ​​of the current pixel a1 and the subsequent pixel a2 are averaged to obtain α1; then the color difference between the next pixel a3 and α1 does not exceed β, then the pixel a3 is added to the monochrome area, and the color values ​​of the three pixels in the monochrome area are averaged to obtain α2, and the color of the subsequent pixel an is continued to be determined. If the color value of the pixel an is equal to the color average value α n-1 If the difference exceeds β, the pixel point an is used as a new reference object and compared with the color value of the subsequent pixel point an+, and the cycle is repeated to find a monochrome area of ​​appropriate size;

[0008] S2. Divide the found monochrome area into squares of N rows and M columns, with each square pixel being a*a. Set the channel color value of the monochrome area to -2 or +2, where -2 represents binary 0 and +2 represents binary 1, to form hidden information of multiple binary numbers.

[0009] Furthermore, before storing the binary information in the monochrome area, a fixed identification information is first stored in the area.

[0010] Furthermore, when extracting information, when the color value of the pixel increases within the threshold range, the binary information changes from 0 to 1, and vice versa, when it decreases, the binary information changes from 1 to 0; if the beginning of the restored binary information is fixed identification information, it means that this area is a monochrome area, and the information carried thereafter is valid hidden data information.

[0011] Beneficial effects of the present invention:

[0012] The present invention does not modify the lowest bit of the color byte, but fine-tunes the color value within a range that is imperceptible to the naked eye, and uses some auxiliary algorithms to ultimately achieve the function of adding a blind watermark to the image. After testing, within a certain range, post-processing of images carrying hidden information, including scaling, rotating, adjusting image brightness, contrast, blurring, sharpening, mosaics, etc., can effectively restore the information hidden in the image.

[0013] The present invention combines the simplicity and clarity of the LSB blind watermarking method, while resisting some post-processing of the image to a certain extent, such as scaling, conversion, blurring, sharpening, adjusting brightness and darkness, etc. It has certain usability and can be used as a supplement to various existing blind watermarking algorithms. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The four squares contained in the monochrome area;

[0015] Figure 2 For the general Figure 1 The effect of the square after the color change;

[0016] Figure 3 This is the effect diagram when the 3*3 pixel square in the grid is reduced to one third; DETAILED DESCRIPTION

[0017] The following is combined with Figure 1-3 The present invention is described in detail:

[0018] The implementation premise of the present invention is to find at least one area in the picture (hereinafter referred to as the original picture) where the information is to be hidden, in which the colors of all pixels are consistent (hereinafter referred to as the monochrome area), and the size of the area depends on the size of the information to be hidden. The information to be hidden can be a monochrome bitmap or a string of binary numbers. With such a monochrome area, we can regularly fine-tune the color of the pixels in the monochrome area based on the picture or binary number to be hidden, so that it carries data information.

[0019] Specifically, the first step is to obtain the monochrome area in the image. If the image contains a pure monochrome area, directly record the monochrome area. Some images themselves contain large monochrome areas, such as a random page of an e-book, which usually contains large white areas. When the height of the monochrome area to be found does not exceed the line spacing of the page text, more such areas will be found. For such images, no fine-tuning is required, and only the monochrome area found on the page needs to be recorded.

[0020] Some pictures, such as blue sky, ocean, animals, and art pictures, may not contain monochrome areas originally, but by fine-tuning the colors of some pixels in the picture, several monochrome areas can be generated. If the picture only contains approximate monochrome areas, the color of the approximate monochrome areas is fine-tuned to generate monochrome areas. The color fine-tuning method is as follows: the first step is to determine the starting position of the monochrome area, and the color difference between the current pixel a1 and the subsequent pixel a2 does not exceed β; the color values ​​of the current pixel a1 and the subsequent pixel a2 are averaged to obtain α1; then the color difference between the next pixel a3 and α1 does not exceed β, then pixel a3 is added to the monochrome area, and the color values ​​of the three pixels in the monochrome area are averaged to obtain α2, and the color of the subsequent pixel an is continued to be determined. If the color value of the pixel an is equal to the color average value α n-1If the difference exceeds β, the pixel point an is used as a new reference object and compared with the color value of the subsequent pixel point an+, and the appropriate size of the monochrome area is found in this cycle. More specifically: search for each pixel point of the picture, take the color of the current pixel point as a reference, if the color value of the subsequent pixel differs from that of the current pixel by no more than a certain value (such as no more than 3), then the current pixel position is first assumed to be the starting position of the monochrome area, and the current color value is averaged with the color value of the subsequent pixel. Then look at the next pixel, if the color value of the pixel does not differ from the previously obtained average value by no more than a certain value, then add the pixel to the previously assumed monochrome area, and at the same time, sum the color value of the pixel with the color values ​​of the previous two pixels (= the color average value * 2 obtained before) and divide it by 3 to obtain a new average value, and continue to judge the color of the subsequent pixel; if the color value of the pixel differs from the color average value of the previous two pixels by more than a certain value, then take the current pixel as a new reference pixel and compare it with the color value of the subsequent pixel. By repeating this process, we can usually find several monochrome areas in the image, record these areas (position, size and other information), and use these monochrome areas to replace the corresponding areas in the original image. The resulting new image can meet the need to store hidden information on it. If such areas cannot be found in the original image, the image cannot store hidden information.

[0021] For example, for a home picture, according to the above method and the set search parameters (specify the size of the monochrome area to be searched as 50*50 pixels), several monochrome areas can be found in the red, green and blue channels of the picture respectively.

[0022] The second step is to modify the color of the monochrome area. The monochrome area is divided into N rows and M columns of squares. The pixel of each square is a*a. The channel color value of the monochrome area is -2 or +2. -2 represents binary 0 and +2 represents binary 1, thus forming the hidden information of the binary number.

[0023] Specifically, after finding the monochrome area in the image, the remaining work is to store the image or binary digital information we want to hide in each monochrome area. For ease of understanding and explanation, we only use the red channel of the monochrome area as an example, but the three channels can be used in actual applications. Suppose you want to store the binary number 0110 in a monochrome area with a red channel color value of 128. If you use the red channel of the first four pixels of the monochrome area and fine-tune its color to make the channel color value -2 or +2 to save the four binary numbers 0110, the image will be vulnerable to various attacks such as image scaling and blurring. So here we choose to divide the monochrome area into a table with 1 row and 4 columns, each table has 8*8 pixels, and then use 0110 as a reference to modify the color of the corresponding 4 small squares, so that the red value of the first and fourth small squares (8*8 pixels) becomes 126, and the red value of the second and third small squares becomes 130. In this way, when the image is scaled, it can be simply understood as the scaling of each small square. Therefore, as long as the scaling multiple of the small square is not large, the information carried by the small square will not be lost. The above method is expressed as Figure 1 to Figure 2 process.

[0024] The method for extracting digital blind watermark is as follows: extracting hidden information in the monochrome area.

[0025] When the image is reduced, the colors of the two adjacent small squares may be averaged, but this does not affect the restoration of hidden information when we extract information. For example, if the width and height of the image are reduced to one-third of the original, a common algorithm is to merge the 3*3 adjacent pixels in the front, back, top and bottom into one pixel by taking the color average. For this example, it means that the red values ​​of pixels P[1]~P[3] and P[4]~P[6] (in the first small square) are averaged, which is still 126; the red values ​​of pixels P[7], P[8], and P[9] (P[7] and P[8] are in the first small square, and P[9] is in the second small square) are averaged, which becomes 127 ( The remainder is rounded off); the red values ​​of pixels P

[10] to P

[12] are averaged, which is still 130; …; the value of the monochrome area before scaling (hiding information) is: 126(*8)130(*8)130(*8)126(*8), where the *8 in the brackets represents 8 consecutive values. After scaling down by 3 times and averaging the pixels, the value of the monochrome area is: 126126127130130…. It can be seen that the pixel values ​​still increase or decrease regularly. Based on the changes in the increase and decrease of pixels, we can restore the hidden binary information.

[0026] Figure 3 The gray square at the top of the image (pixel 3*3) will become a Figure 3 The pixel indicated by the bottom line.

[0027] For other image adjustments, such as blurring and adjusting color tone, the analysis methods are basically the same. Due to space limitations, they will not be described one by one here.

[0028] When extracting information, when the color value of a pixel rises within the threshold range, it means that the hidden binary information of the pixels before and after this point has changed from 0 to 1. Conversely, it has decreased from 1 to 0.

[0029] If the monochrome area in the original image is identified, it is also a problem that must be faced when extracting hidden information. The solution is that before storing binary information in the monochrome area, we first store a fixed identification information, such as 1010. When identifying the hidden information of the image, we still follow the previous method and restore the binary information according to the change of pixel color. If the beginning of the restored binary information is 1010, it means that this area is a monochrome area, and the information carried afterwards is valid hidden data information. It should be emphasized that when the pixel change value exceeds the threshold, such as the value of the red channel of the two pixels before and after changes from 128 to 142, or the color value of the adjacent pixels has been increasing, from 128 to 142, this means that this area is not a monochrome area, so even if a value such as 1010 can be obtained according to the rise and fall of the color value, it cannot be used as the start mark of the monochrome area. It is not troublesome to identify this situation. We only need to record the starting value of a color. If the color value of the pixel differs from the recorded color starting value by more than a certain value, it means that this area is not a monochrome area.

Claims

1. A method for producing a digital blind watermark, characterized in that: It includes the following steps: S1. Obtain the monochrome area in the image. If the image contains a pure monochrome area, directly record the monochrome area; if the image only contains an approximate monochrome area, fine-tune the color of the approximate monochrome area to produce a monochrome area. The color fine-tuning method is as follows: the first step is to determine the starting position of the monochrome area. The color difference between the current pixel a1 and the subsequent pixel a2 does not exceed β; the color values ​​of the current pixel a1 and the subsequent pixel a2 are averaged to obtain α1; then the color difference between the next pixel a3 and α1 does not exceed β, then the pixel a3 is added to the monochrome area, and the color values ​​of the three pixels in the monochrome area are averaged to obtain α2, and the color of the subsequent pixel an is continued to be determined. If the color value of the pixel an is equal to the color average value α n-1 If the difference exceeds β, the pixel point an is used as a new reference object and compared with the color value of the subsequent pixel point an+, and the cycle is repeated to find a monochrome area of ​​appropriate size; S2. Divide the found monochrome area into squares of N rows and M columns, with each square pixel being a*a. Set the channel color value of the monochrome area to -2 or +2, where -2 represents binary 0 and +2 represents binary 1, to form hidden information of multiple binary numbers.

2. A method for producing a digital blind watermark according to claim 1, characterized in that: Before storing binary information in the monochrome area, a fixed identification information is first stored in the area.

3. A digital blind watermark extraction method, using the digital blind watermark production method according to claim 1, characterized in that: When extracting information, when the color value of the pixel increases within the threshold range, the binary information changes from 0 to 1. Conversely, when it decreases, the binary information changes from 1 to 0. If the beginning of the restored binary information is fixed identification information, it means that this area is a monochrome area, and the information carried afterwards is valid hidden data information.

Images