TW201333873A - Digital image authentication method - Google Patents

Abstract

The present invention provides a digital image authentication method, which includes the steps of: performing a column inversion arrangement to plural significant bits of plural pixels; generating parity check bits for each pixel based on a Hamming code; calculating corresponding position for each pixel based on a Torus automorphism; performing a bit rotation to the parity check bits; hiding the parity check bits into the corresponding position of each pixel; determining whether the parity check bits of each modified pixel are modified or not; determining at least one modified pixel; and predicting the modified pixel based on the parity check bits. With the digital image authentication method disclosed by the present invention, it is able to detect and effectively recover the modified complex area.

Description

Digital image authentication method

The invention relates to a digital image authentication method, in particular to a digital image authentication method capable of detecting and effectively replying to a modified complex area.

Due to the development of digitalization, digital images have become easy to modify, and anyone can modify digital images through image processing software. Therefore, the security mechanism of digital images and the demand for intellectual property rights have become more and more urgent. One of the techniques developed in the art is called Image Authentication, which is designed to ensure the integrity of digital image content. Since the purpose is to ensure integrity, it must be capable of detecting whether the digital image has been tampered with. What is more, the modified area can be located and further restored to an unmodified state.

Image Authentication technology can be practiced in two broad categories: (1) Digital Signature Approach image authentication technology and (2) Watermark-based Approach image authentication technology. .

The digital signature technology of the Digital Signature Approach is based on extracting the feature information of the digital image. After simplifying the feature message, the simplified feature information is saved separately as the authentication data (Authentication). Data). When it is necessary to confirm whether the digital image has been modified, the verifier must provide the authentication data in order to authenticate the digital image. Ahmed et al. proposed this type of image authentication technology in 2010. The method first uses the secret key to take the remainder of the pixel (Modulu) as an image feature. Then, the image features are hashed to obtain the hash value. Finally, the hash value is quantified to obtain the image authentication data. When it is required to detect a digital image, the authentication data can be used to locate the modified portion of the digital image. Since the authentication data is independent of the digital image, any modification of the digital image will not be modified to the authentication data. However, the downside of this type of technology is that the owner must carefully save the certification data for future certification.

On the other hand, the watermark-based approach image authentication technology, after extracting and simplifying the feature information of the digital image, is hidden back into the original digital image to generate the authentication code. Watermarked Image. For the purpose of authentication, the authentication data must be extracted from the digital image that has been modified, and the authentication data is taken out to detect whether the image has been modified. Such technologies can be broken down into two categories based on their ability to recover modified areas. The first type of technology only has the ability to perform the positioning action of the modified area from the obtained authentication data. For example, in 2010, Patra et al. proposed using Chinese Remainder Theorem (CRT) on the coefficients of a Discrete Cosine Transform (DCT) block to generate authentication data. The generated authentication data will be hidden back into the digital image block. According to the experimental results, the method of Patra et al. does not cause excessive damage to the impact after the identification information is hidden. In addition, images of certified data will not be misidentified as malicious modifications after being processed by some standard image processing techniques (such as JPEG compression). However, the amount of information that can be possessed by such technologies is very limited. The reason is that the space in which the data is to be hidden must be processed by standard image processing technology to ensure that the data stored is not modified. Space in the image is inherently very limited. Therefore, this kind of technology can only detect and locate the modified range, and the space in which the authentication code can be hidden is too much to hide the reply data together, so this kind of technology cannot be modified. The area is back to recovery.

In order to be able to recover the modified area back, a second type of technology was proposed. Most of these technologies use some of the techniques of image compression to generate authentication data and hide it in digital images. When the digital image is modified, the image compression code of the area can be taken out from the authentication data, and the modified area is restored.

The technology of this kind is as follows: Yang and Shen's method proposed in 2010 to first perform digital image compression (Vector Quantization (VQ)) image compression technology to generate an image index table (Index Table). The Index Table is used as information to be used to restore digital images in the future, and this reply information will be used as authentication data and hidden back into the digital image. This uses image compression techniques to generate the authentication code and uses the authentication code to recover the modified region. The fineness of the modified region that can be recovered in the future is limited by the image compression method used. Taking the method of Yang and Shen as an example, since the method uses vector coding, the digital image is compressed and an authentication code is generated. Therefore, when the modified region is to be restored, the granularity of the reply is 44 pixels, because the vector coding method performs vector coding by using 44 pixels as a block unit. In the same way, the method proposed by Lee and Lin in 2008 is to use 22 pixel blocks as a coding unit. The method firstly finds the average of the pixels of the block, and retains the average value before converting to the binary code. The five most important bits, combined with the first five most significant bits of the average of the pixels of other blocks, produce authentication data that is hidden back into the block. Obviously, Lee and Lin's method response has a fineness of 22 pixels when the modified region is to be restored.

Therefore, how to solve the above problems and deficiencies in the above-mentioned applications, that is, the inventors of the present invention and those involved in the industry are eager to study the direction of improvement.

Therefore, in view of the above-mentioned deficiencies, the inventors of the present invention have collected relevant materials, evaluated and considered them through multiple parties, and have been designing such digital image authentication methods through continuous trial and modification through years of experience in the industry. Patent holder.

The main object of the present invention is to provide a digital image authentication method that can detect and effectively reply to a modified complex area.

In order to achieve the above object, a digital image authentication method of the present invention comprises the following steps:

(110) providing a digital image having a plurality of pixels;

(120) performing reverse alignment on the plurality of significant bits of each pixel;

(130) generating a parity check code for each pixel based on a Hamming coding method;

(140) calculating a corresponding position of each pixel based on a circular symmetric algorithm;

(150) performing bit rotation on the parity check code of each pixel;

(160) hiding the parity check code of each pixel into a corresponding position of each pixel;

(170) determining whether the parity check code of each modified pixel is modified;

(180) determining at least one modified pixel;

(190) predicting the modified pixel based on the parity check code.

In a preferred embodiment, the one-bit check code is generated for each pixel based on a Hamming coding method, and a three-bit parity check code is generated for four bits in each pixel.

In a preferred embodiment, the parity check code of each pixel is hidden in a corresponding position of each pixel, and further includes hiding an indicator bit in a corresponding position of each pixel.

In a preferred embodiment, the bit rotation further includes a bit position J' of the new parity check code, a bit position J of the original parity check code, a randomly generated i-th number R _i, and a rotation The total number of bits M is rotated by J '=( J + R _i )mod M .

In a preferred embodiment, the predicting the modified pixel based on the parity check code further includes the following steps:

(191) generating a matrix for calculating the number of unmodified pixels around the modified pixel;

(192) restoring the largest number of unmodified pixels around the modified pixel;

(193) determining whether the modified pixel is present;

(194) If the modified pixel is present, the step (191) is repeatedly performed.

In a preferred embodiment, the matrix size is the same as the digital image.

Accordingly, the present invention proposes a complex region in which authentication and response of digital images are modified based on a Hamming Code. First, the present invention first reversely arranges the complex important bits of each pixel, and then generates a parity check bit (Parity Check Bits) for each pixel based on the Hamming coding method, and for the sake of security, the parity The meta-check code will be processed by a Bit rotation technique to generate the rotated parity check code, which is hidden in the Torus automorphism algorithm. The corresponding position of the calculation, since the parity check code is generated from a single pixel by using the Hamming coding method, the restored fineness is 1 pixel when the modified pixel is restored compared to the prior art. Also, due to the special characteristics of the Hamming coding method, when the authenticated digital image is transmitted over the network and the Burst Bit Errors are generated, the characteristics of the Hamming coding method can be recovered.

When it is necessary to restore the modified pixel, the modified pixel can be predicted based on the parity check code. According to the experimental results, the present invention can generate a high-quality authenticated image, and has a very high High resilience.

In order to achieve the above objects and effects, the technical means and the structure of the present invention will be described in detail with reference to the preferred embodiments of the present invention.

Referring to the first embodiment, which is a flowchart of a preferred embodiment of the present invention, it can be clearly seen from the figure that the digital image authentication method of the present invention comprises the following steps:

(110) providing a digital image having a plurality of pixels;

(120) performing reverse alignment on the plurality of significant bits of each pixel;

(130) generating a parity check code for each pixel based on a Hamming coding method;

(140) calculating a corresponding position of each pixel based on a circular symmetric algorithm;

(150) performing bit rotation on the parity check code of each pixel;

(160) hiding the parity check code of each pixel into a corresponding position of each pixel;

(170) determining whether the parity check code of each modified pixel is modified;

(180) determining at least one modified pixel;

(190) predicting the modified pixel based on the parity check code

In this step (110), the pixel is the smallest unit constituting the digital image 1. It should be easily understood by those having ordinary knowledge in the art. In this embodiment, the pixel is set to 8 bits (bits). ), if it is 24-bit or 32-bit, it is also a feasible solution.

In the steps (120) and (130), the one-bit check code is generated for each pixel based on a Hamming coding method, and the three-bit parity check is generated for four bits in each pixel. code.

The invention uses the Hamming coding method to achieve the purpose of image authentication. Before explaining the present invention, the present invention first describes the Hamming coding method. The Hamming coding method is an encoding method, and the receiving end can correct the error that may be generated according to the received encoding itself. In order to be able to correct the Hamming code, the original data must be used to generate additional data, and additional information is attached to the original data. The basic concept of the Hamming coding method is "Parity check", which means that the Hamming coding method uses additional data to authenticate the original data. The amount of additional data attached is determined by the Hamming inequality rule. The present invention does not describe the principle of this inequality here, and those skilled in the art should be able to easily understand it. According to this inequality, the original data of four bits requires three bits of additional data, so the encoded data becomes seven bits in total, which is called "(7,4) Hamming code".

Now I will show you how to use the "(7,4) Hamming code" to generate the parity check code. Please also refer to the third figure. Let there be four bits of raw data ( D ₁ , D ₂ , D ₃ , D ₄ ), and the other three bits ( P ₁ , P ₂ , P ₃ ) are parity check codes, and each round will be wrapped. The original data of the relevant company and the certification data, and the number of bits of the related information, the total number of the values of "1" must be an even number. In order to satisfy the above conditions, the inspection data bits can be uniquely determined for the original data of any four bits. That is, the value of the parity check code can be calculated such that the number of bit values in each circle is "1" and the number is even. For example, the parity check code bit P _{1 is associated} with the original data bits D ₁ , D _{2 ,} and D ₄ . If the values of D ₁ , D _{2 ,} and D ₄ are 1, 0, and 1, respectively, the parity check bit P ₁ must be 0 in order to make the total number of bits in the circle having a value of "1" an even number. This is enough to meet the requirements. Others, such as bits P ₁ and P _{2 ,} also use the same decision mode to determine their values, so that a three-bit parity check code can be generated.

In the method of the present invention, however, the three bit authentication material does not come directly from the most significant four bits of the pixel. The most important four bits, that is, the important bits must be reversed, and then three bits of authentication data are generated. Please also refer to the second figure. The reason why the bit must be reversed is that after the bit is inverted, we can know the value of the most significant bit in the original pixel based on the value of the three bit check code. The principle can be explained by the third figure. Please also refer to the third figure. Since the number of bit values in each circle is even, the two sets of data bits should have the same When the parity check code, the values of D ₁ , D ₂ and D ₃ of the two sets of data bits must be completely different, but the values of D ₄ must be the same, so that they can have the same parity check code. In the example of the third figure, the two data bits (1011) ₂ and (0101) ₂ will produce the same parity check code (010) ₂ . It can be observed that the value of the same bit D ₄ is 1. More precisely, regardless of how the data bits are different, when they have the same parity check code, their data bits D ₄ must be the same, both at 1 or both. So as long as we place the most significant bit in the D ₄ position, we can know the value of the most significant bit of the pixel (that is, D ₄ ) based on the value of the parity check code. In this way, more accurate prediction of pixel values can be made, so in the method of the present invention, the most important four bits are directly inverted, so that the most important bits can be placed at the D ₄ position. The aligned data bits are inverted to generate a parity check code.

After generating the parity check code, the generated parity check data is hidden back into three less important bits in the "original" pixel. This action seems to be very reasonable, but the purpose of the present invention is not only to correct a bit error, but also to modify and recover the entire area. If the parity check code is directly hidden back into the "original" pixel, once the pixel is in the modified area, the important bit data of the pixel and the parity check code are simultaneously destroyed.

In this step (140), in order to solve this problem, please refer to the fourth figure at the same time, the present invention hides the parity check code generated by the pixel A to the less important bit in the other pixel B, for any One pixel A, how we decide which pixel represents pixel B is very important. Because when we want to do authentication, we must extract the authentication data. If the pixel B was originally found, the pixel B could not be found when the data was retrieved. Of course, the authentication data could not be extracted. Therefore, we must find the corresponding pixel B for any one pixel A, and any two pixels, and can not find the same pixel to store data. Torus automorphism has this function. This algorithm was proposed in 1996 to disrupt the values in two-dimensional space. The formula is as follows:

Wherein x _i and y _i represent the i-th bit image pixels, located on his row y _i x _i columns. Further aspect of the x _i 'and y _i' represents the i-th pixel corresponding to its new position, which is located on a position x _i 'row y _i' column. The number of pixels in the length and width of the digital image is N , and the value of k can be regarded as a key. According to this key, each pixel can be calculated from its current position coordinates to calculate the coordinates of the new position.

For example, please refer to the fourth figure at the same time, assuming that the value of N is 9, the value of k is 3, the value of pixel A is 176, its position is (1, 1), and the value 176 is represented by binary. When the most important 4 bits are (1011) ₂ , after the rearrangement, (1101) _{2 is} generated. From these 4 bits, the parity check data can be generated as (100) ₂ , according to the circular symmetric calculus. From the old position coordinates, the new coordinates can be calculated as (2, 7), that is, the position of the pixel B displayed on the fourth picture, so we can hide the parity check data to the pixel B. The value of the original pixel B is 47, and the binary representation is (00101111) ₂ . We replace the least significant three bits in pixel B with the parity check code generated by pixel A, so the value of pixel B becomes (00101 100 ) ₂ . Of course, the least significant three bits in pixel A will also be used to check the parity of the hidden pixels. Taking the fourth figure as an example, pixel C will hide its parity check data in pixel A.

In-depth study of the effect of the circular symmetric algorithm on the present invention, we can find that the k value is still limited by N , more precisely, the value of k will be between 1 and N -1. Since we can calculate the parity check code for each pixel, we can test the value of k from 1 to N -1 to confirm whether the last three bits of the new position tested are the same checksum code. Then after a maximum of N -1 tests, you can find the value of k . Since the k value can be found, the forger can modify the graph arbitrarily, and modify the pixel, calculate a new parity check code, and then calculate the position and check the new parity according to the circular symmetric algorithm and the value of k . The code is hidden, so that the modified part cannot be pointed out when detecting.

In this step (150) and (160), in order to solve this problem, before the parity check code is concealed, the bit rotation technique must be used first, and the bit rotation technique is used to rotate the bits of the parity check code. The position of the element, in which the invention must use a second key k ', using k ' to generate a series of random numbers R ₁ , R ₂ , ..., R _{N × N} . When the parity check code is hidden in the pixel, the bit position of the parity check code is rotated according to the randomly generated numbers, and the rotation is as follows:

J '=( J + R _i )mod M

Wherein R _i will be designated as the i-th, the digital random number to be used in the i-th pixel. J represents the bit position of the original parity check data, J ' represents the bit position of the new parity check data, and M represents the total number of bits to be rotated.

For example, please refer to the fourth and fifth figures at the same time, the number of bits of the parity check code is 3, so the value of M is 3. In the example, the value to be hidden in pixel B is (100) ₂ . Assuming that the random number assigned to this pixel is 121, the result after the bit is rotated, the rotated bit is (010) ₂ , This bit is hidden back into pixel B, and the last pixel value will become (00101010) ₂ .

Furthermore, in this step (160), the parity check code of each pixel is hidden in a corresponding position of each pixel, and further includes hiding an indicator bit in a corresponding position of each pixel. When recovering a modified pixel, the present invention predicts the four most significant bit values of the modified pixel based on the value of the parity check code. Referring to Figure 7, you can see that each parity check code will correspond to two possible bit data values. The detailed predictions will be mentioned later. But when we need to further increase the correctness of the restored pixels, we can add one more bit to indicate which of the two possible data bits for the case that is easier to misjudge. As for what is easier to misjudge? Please refer to the seventh figure. When the value of the parity check code is 3 or 4, the difference between the two data bit values will get the minimum difference of 1 (please refer to the last field in the seventh figure), so It is a situation that is more likely to cause misjudgment. Although this situation is prone to misjudgment, it does not cause the pixel values to differ a lot because their minimum difference is 1. However, when the situation becomes more prone to misjudgment and the result of the misjudgment results in a large difference in the value of the pixel, the present invention adds an "indicator bit" to indicate which of the two possible data bits, that is, These will produce false positive pixels, and a total of 4 bits (3 bit parity check code and one bit indicator bit) will be hidden into the pixel.

In the method of the present invention, although the value of the parity check code is 3 or 4, the misjudgment is most likely to occur, but the difference is only 1, which means that even if the error is judged, the value difference of the pixel is not caused much. Therefore, this situation does not add an "indicator bit". In contrast, the present invention adds one more indicator bit when the value of the parity check code is 2 or 5. The method of the present name is " method #1 of the indicator bit ." In this way, a more accurate prediction can be achieved. When we have a higher demand for the quality of the restoration, in addition to the above-mentioned pixel with the parity check code value of 2 or 5, plus an "indicator bit", it is also possible to increase the value of the parity check code to 1 or 6. Pixels, plus an "indicator bit" to achieve higher recovery correctness, we call this method " method #2 with indicator bits ". Finally, if "indicator bits" are added to all pixels, we call this method " method #3 with indicator bits ".

In this step (170), the image change may be caused by deliberate or unintentional modification. When the image is transmitted over the network, it may be affected by poor network transmission quality, and unintentional modification may occur. Unintentional modification produces only one bit error, and this error occurs in the most significant 4 bits or the least significant three bits, which can be found according to the Hamming code encoding method, and Correct this error back.

For deliberate modification, we must first divide the modified pixel area to be modified. In (Figure 6), the gray part is assumed to be the modified part. We call this area X, where the value of pixel A has been edited. As described above, the parity check data of the pixel A is hidden in the pixel B, and the parity check data of the pixel C is hidden in A. Since the pixel A has been modified, the value of the parity check data calculated after the modification will not be the same as the value of the parity check data previously hidden in the pixel B. Therefore, both the pixel A and the pixel B will be determined as Modified pixels. Similarly, pixel C is also judged to be modified because its parity check data is hidden in pixel A, and pixel A has been modified. Therefore, we can test all the pixels of the entire image, and mark the pixels that are determined to be modified. Finally, we can see that most of the pixels in the X area will be marked with modified marks. In addition, some individual pixels are marked with modified symbols, such as pixel B and pixel C. It can be seen from the observation that marking B and pixel C with the modified mark is the result of misjudgment.

In this step (180), we must remove the modified token for the misjudged pixel. Here, we use the Morphological operations to achieve our goal. The operator contains two parts: the erosion operator (Erosion). Operations) and the growth operator (Dilation operations), after using the erosion operator and the growth operator in sequence, the independent marker points or the area containing only a few marker points will be eliminated, and finally only in the X region The modified pixels will be marked with a modified token. After the detection process causes some pixels to be marked with a modified mark, the next step of the present invention is to try to recover the pixels of the marked modified mark. Since the pixel values in region X have been modified, there will be no messages that help restore these pixels. However, the parity check codes for these pixels are not necessarily modified, so we will discuss how to restore these modifications. Pixel.

In this step (190), first, if the parity check code is not modified and it has an "indicator bit", such pixels are preferentially restored because their values can be uniquely determined. If only the parity check code exists and has not been modified, we can know the two possible data bits of the most important four bits of the original pixel according to the value of the parity check code, as shown in the seventh figure. In addition, based on the parity check code, the bit value of the most significant bit of the predicted pixel can be known. For example, when we know that a modified pixel has a parity check code of (000) ₂ , we can refer to the seventh figure that its data bit has a possible value of (0000) ₂ or (0111) ₂ , so When replying to a pixel, one of the two possible data bits must be selected as the predicted data bit to recover the modified pixel. It is worth noting that, according to the parity check code, we already know that the data bit to be predicted has the most significant bit value of zero.

In the step (190), the following steps are further included:

(191) generating a matrix for calculating the number of unmodified pixels around the modified pixel;

(192) restoring the largest number of unmodified pixels around the modified pixel;

(193) determining whether the modified pixel is present;

(194) If the modified pixel is present, the step (191) is repeatedly performed.

In order to select one of the two possible data bits, we additionally declare a matrix M of the same size as the digital image, as shown in the eighth figure, to calculate how many pixels around the modified pixel are not Modified (contains pixels that have been restored). The value of the element in the matrix M is how many pixels around the recorded pixel are marked as "unmodified pixels" (including pixels that have been recovered). When the elements in the matrix have larger values, there are more pixels around them that can be referenced, so these modified pixels with larger values will be restored first.

Please refer to the eighth figure, because the maximum value of the elements in the matrix is 6, the corresponding pixels will be restored first. Since the value of the parity check code is already known, it is possible to know the most significant bit value of the modified pixel 11. When referring to the surrounding "unmodified pixels 12", only pixels with the same most significant bit value can participate in the operation to find the average of these pixels. Through the average, two possible data bits can be calculated, which is the closest to the average, and the data bit closest to the average is used to recover the four most significant bits of the modified pixel 11.

Each round, we only recover the modified pixel 11 with the largest value. The recovered pixels are then marked as "unmodified pixels 12", and then the element values in matrix M are recalculated, and the corresponding pixels in the matrix M having the largest value are resumed. Taking the eighth figure as an example, in the first round, all pixels corresponding to the value of 6 are restored, and the element values in the matrix M are recalculated. In the second round, we look for the element with the largest value in the matrix (that is, the pixel of value 6) to perform the recovery action until all the modified pixels 11 are marked as "unmodified pixels 12".

In summary, the present invention uses Hamming coding to achieve image authentication. In the present embodiment, three programs are mainly included. The first program is called a hidden program, that is, an authentication code is generated and hidden in a digital image, as in steps (110) to (160). The second program is the detection program, which is how to divide the modified area when the image is modified, as in steps (170) ~ (180). The last program is the recovery program, that is, how to restore the area determined to be modified, as in step (190).

The experimental method of the digital image authentication method of the present invention and the result thereof will be described below. In the present embodiment, a digital image of 512 512 pixels called Lena is used as an experimental object, and Lena is a digital image file. Image files have long been widely used in image compression and processing results testing, and those skilled in the art should be able to easily understand them, as shown in the ninth figure. Lena was chosen as the experimental image because this digital image contains many smooth and complex areas that fully demonstrate the superiority of the method of the present invention.

First, the present invention generates a parity check code for each pixel of the experimental digital image, and hides it into other pixels to obtain an image of the stored authentication data, and its image quality is shown in the tenth (a)-(d) . The tenth figure (a) is the authentication image generated by the " original proposed method ", that is, the method of not adding the indicator bit, and the tenth figure (b)-(d) is respectively using the " join indication". The authentication method generated by the bit method #1 ”, “ method #2 of adding indicator bit ” and “ method #3 of adding indicator bit ”. Compared with the calculation of image quality, the present invention uses PSNR (Peak-Signal-to-Noise Ratio) as a criterion for calculating the image quality of the stored authentication data. The formula for calculating PSNR is listed below:

The variables H and W represent the high-heel width of the image, I ( j , k ) represents the pixel value of the k- th row of the j-th column of the original image, and I' ( j , k ) represents the image of the stored authentication data. The pixel value of the kth line of j column.

The first experiment of the present invention is to demonstrate the ability of the method of the present invention to detect and recover from sudden bit errors. First, 5,000 bits are randomly selected from the authenticated electronic image, and the values of these bits are inverted. , that is, 0 becomes 1 or 1 becomes 0, and after a burst bit error, a part of the generated electronic image is displayed in the eleventh figure (a), and after the recovery procedure according to the present invention, The result is shown in Figure 11 (b).

A second experiment of the present invention is to demonstrate the ability of the method of the present invention to detect and recover maliciously modified electronic images. First, first select four areas from the certified electronic image for modification. The first area is a smooth area, and the second area is a complex area. The other two areas contain both smooth and complex. The area (Zone I and Area II) is modified as shown in Figure 12.

For the " original proposed method ", " method #1 for adding indicator bits ", " method #2 for adding indicator bits " and " method #3 for adding indicator bits ", the authentication image is generated by the tenth The position shown in the two figures is modified, and the modified pixels are restored by the detection recovery method of the present invention. The recovery result for the modified pixels of the smooth region is shown in the thirteenth figure in the case of different methods. The recovery results for the modified pixels of the complex region are shown in the fourteenth figure in the case of different methods. The fifteenth and sixteenth graphs show the recovery results for modified pixels having both smooth and complex regions in the case of different methods. Through the experimental results, it can be seen that most of the pixels are successfully restored, and the smooth background area or the complex hair texture can be successfully restored. Moreover, for some areas such as the eye, the method of the present invention can detect and recover the modified area very accurately. It can also be found that as the amount of pixels with the "indicator bit" increases, the effect of recovery is more significant. However, as the amount of pixels added with the "indicator bit" increases, the image value decreases, as shown in the tenth figure.

Referring to the drawings, the present invention has the following advantages over conventional techniques:

In the value of digital image transactions, it must be able to ensure that the purchaser can verify the integrity of the image. Such digital media transactions on the Internet have become more and more popular, and the present invention can surely maintain the integrity of the digital content.

Through the above detailed description, it can fully demonstrate that the object and effect of the present invention are both progressive in implementation, highly industrially usable, and are new inventions not previously seen on the market, and fully comply with the invention patent requirements. , 提出 apply in accordance with the law. The above is only the preferred embodiment of the present invention, and is not intended to limit the scope of the invention. All changes and modifications made in accordance with the scope of the invention shall fall within the scope covered by the patent of the invention. I would like to ask your review committee to give a clear explanation and pray for it.

(110)~(190). . . step

1. . . image

11. . . Modified pixel

12. . . Unmodified pixel

D1~D4. . . Data bit

P1~P3. . . Parity check code

A, B, C. . . Pixel

X. . . region

M. . . matrix

The first figure is a flow chart of a preferred embodiment of the invention.

The second figure is a first schematic diagram of the implementation of the preferred embodiment of the present invention. The four data bits of the pixel are reversely arranged and named as D ₁ to D ₄ , and the D ₁ to D ₄ data bits are used to generate P _{1 .} Go to the P ₃ parity check data.

The third figure is a schematic diagram of the implementation of the preferred embodiment of the present invention. It is illustrated that D ₁ to D ₄ are four data bits and three identical parity check data generated by P ₁ to P ₃ .

The fourth figure is a third embodiment of the preferred embodiment of the present invention, illustrating an electronic image of size 9 9 and A, B, and C for the three pixels above the image.

The fifth figure is a schematic diagram of the implementation of the preferred embodiment of the present invention. The description of J indicates the bit position of the original parity check data, and J' indicates the bit position of the new parity check data.

The sixth figure is a schematic diagram of the implementation of the preferred embodiment of the present invention. A, B, and C are three pixels on the image, and the area X (gray portion) is subjected to deliberate modification.

Figure 7 is a schematic diagram of the implementation of the preferred embodiment of the present invention, illustrating that the most significant bit refers to the first bit, that is, the bit framed by the box, and the last column is two original data bits (also Is the difference between the values of the first field).

The eighth figure is a schematic diagram of the implementation of the preferred embodiment of the present invention. The size of the matrix M is the same as the size of the original image. The range of the black frame is the modified area, and the number represents how many pixels in the vicinity of the pixel are “unmodified pixels” (or The total number of pixels that have been restored. .

The ninth diagram is a schematic diagram of the implementation of the preferred embodiment of the present invention, which is an experimental digital image.

The tenth figure is a schematic diagram IX of the implementation of the preferred embodiment of the present invention, illustrating an electronic image of the possessed authentication data.

(a) The authentication code is hidden by the "original method" and its image quality is PSPR = 37.57 .

(b) The authentication code is hidden by "Method #1 of adding the indicator bit" , and the image quality is PSPR = 35.11.

(c) The authentication code is hidden by "Method #2 of adding the indicator bit" , and the image quality is PSPR = 32.99.

(d) The authentication code is hidden by "Method #3 of adding the indicator bit" , and the image quality is PSPR = 31.96.

11 is a schematic diagram of the implementation of a preferred embodiment of the present invention, illustrating the ability of the present invention to resist sudden bit errors.

(a) For the eleventh figure (a), an electronic image of a sudden bit error (a random exchange of 0 and 1 for a value of 5000 bits).

(b) An electronic image restored by the present invention.

The twelfth embodiment is a schematic diagram of the implementation of the preferred embodiment of the present invention. The invention is modified to modify the image to which the authentication material has been added.

(a) Modify the smooth area.

(b) Modifications to complex areas.

(c) Modifying the area I with smooth and complex areas.

(d) Modifying Area II with smooth and complex areas.

Figure 13 is a schematic view of the preferred embodiment of the present invention, illustrating the ability of the various methods of the present invention to resist malicious modifications to smooth areas.

(a) Restoring the modified area with the " original proposed method ".

(b) Restore the modified area with " Method #1 of joining indicator ".

(c) Restore the modified area with " Method #2 of joining indicator ".

(d) Restoring the modified area with " Method #3 of joining indication bits ".

Figure 14 is a schematic diagram of the implementation of a preferred embodiment of the present invention, illustrating the ability of various methods of the present invention to resist malicious modifications in complex areas.

(a) Restoring the modified area with the " original proposed method ".

(b) Restore the modified area with " Method #1 of joining indicator ".

(c) Restore the modified area with " Method #2 of joining indicator ".

(d) Restoring the modified area with "Method #3 of joining the indication bit" .

The fifteenth embodiment is a schematic diagram of the implementation of the preferred embodiment of the present invention, illustrating the ability of the various methods of the present invention to be maliciously modified against areas I having smooth and complex regions.

(a) Restoring the modified area with the "original proposed method" .

(b) Restoring the modified area with "Method #1 of joining the indication bit" .

(c) Restoring the modified area with "Method #2 of joining the indication bit" .

(d) Restoring the modified area with "Method #3 of joining the indication bit" .

Figure 16 is a schematic diagram showing the implementation of a preferred embodiment of the present invention. The method of the present invention demonstrates the ability of the region II having smooth and complex regions to be maliciously modified.

(a) Restoring the modified area with the "original proposed method" .

(b) Restoring the modified area with "Method #1 of joining the indication bit" .

(c) Restoring the modified area with "Method #2 of joining the indication bit" .

(d) Restoring the modified area with "Method #3 of joining the indication bit" .

(110)~(190). . . step

Claims (7)

A digital image authentication method includes the following steps: (110) providing a digital image having a plurality of pixels; (120) inverting the plurality of important bits of each pixel; (130) based on a Hamming code The method generates a parity check code for each pixel; (140) calculating a corresponding position of each pixel based on a circular symmetric algorithm; (150) performing bit rotation on the same check code of each pixel; (160) hiding the parity check code of each pixel into a corresponding position of each pixel; (170) determining whether the parity check code of each modified pixel is modified; (180) determining at least one modified pixel And (190) predicting the modified pixel based on the parity check code. The digital image authentication method according to claim 1, wherein the one-bit check code is generated for each pixel based on a Hamming coding method, and three bits are generated for four bits in each pixel. The parity check code. The digital image authentication method of claim 1, wherein the parity check code of each pixel is hidden in a corresponding position of each pixel, and further comprises hiding an indicator bit into the phase of each pixel. Corresponding location. The digital image authentication method according to claim 1, wherein the bit rotation further includes a bit position J′ of the new parity check code, a bit position J of the original parity check code, and a randomly generated bit. The number i of the i numbers R _i and the total number of bits to be rotated is J '=( J + R _i ) mod M . The digital image authentication method according to claim 1, wherein the step of determining that the at least one modified pixel is based on an erosion operator and a growth operator to determine the at least one modified pixel. The digital image authentication method according to claim 1, wherein the step of predicting the modified pixel based on the parity check code further comprises the following steps: (191) generating a matrix for calculating that the modified pixel is not modified. The number of pixels; (192) restores the largest number of unmodified pixels around the modified pixel; (193) determines whether the modified pixel is present; and (194) if the modified pixel is present, repeats the step (191). The digital image authentication method according to claim 6, wherein the matrix size is the same as the digital image.