Image blind watermarking method based on exponential function

Abstract

The invention belongs to the field of digital watermarking technology and information security, and relates to an image blind watermarking method based on an exponential function. In most existing watermark embedding methods based on singular value decomposition, watermark information is often embedded through an additive method, and original watermarks or host information is often needed in the watermark extraction process. Based on the above problems, the invention provides a blind watermarking method based on an exponential function, and by means of the characteristics of the exponential function, a watermark and a host are respectively used as a cardinal number and an exponent of the exponential function, so that a carrier image containing the watermark can be obtained. In the extraction process, the watermark can be recovered by performing absolute value operation on the carrier image. The whole extraction process avoids the participation of original watermarks and hosts, realizes the blind extraction of the watermarks, and resists image processing attacks at the same time. The figure of the abstract is an embedded flow chart of the image blind watermarking method based on the exponential function, and the method is mainly applied to image copyright protection and information security.

Description

Technical Field

[0001] The invention relates to an exponential function embedding method, in particular to an image blind watermarking method based on an exponential function, and belongs to the field of image copyright protection and information security. Background Art

[0002] As the main means of protecting image information security, image watermarking technology has been widely studied in recent years and has become one of the research hotspots in the field of image information security. Image watermarking is to hide information, so that the content of the image is hidden and cannot be visually detected, so as to protect the information security of the image. According to the visibility of the watermark, the image watermarking method can be divided into visible watermarking method and invisible watermarking method. Visible watermark is a watermark that can be directly seen by the human eye and is mainly used in images and videos. Visible watermarking technology prevents information from being spread on the Internet and misused by others by adding semi-transparent text or logos to documents, pictures and videos. The main feature of visible watermark is that the watermark is visible but not conspicuous in the carrier, and the visible watermark is usually difficult to remove without affecting the quality of the carrier, so as to achieve copyright protection of the carrier. However, with the rapid development of digital image processing technology, attackers can use a variety of software technologies and image processing technologies to erase the visible watermark from the carrier without a trace, which also poses a great security risk to visible watermarking technology.

[0003] Corresponding to the visible watermark is a more secure invisible watermark technology, that is, a watermark that cannot be directly seen by the human eye. The invisible watermark technology has a wider range of applications, including copyright protection of text, images, audio and video. The main feature of the invisible watermark is that the watermark is visually imperceptible. The copyright owner can embed the identity information representing himself as a watermark into the carrier. When a copyright dispute occurs, it is only necessary to extract the watermark information to verify the copyright ownership of the item. Therefore, the invisible digital watermark technology has special practical significance in the application of item anti-counterfeiting.

[0004] In most of the existing digital watermark embedding methods, watermark information is embedded into the host image mainly through additive watermarking. In the watermark extraction process, most of the existing watermarking methods based on singular value decomposition often require the participation of the original watermark image or the original host image, which makes the watermark system prone to false positive problems and vulnerable to image processing attacks.

[0005] Therefore, the development, design and research of watermark systems have become the focus and important research direction in the field of information security. Proposing a safe and effective watermark embedding algorithm to solve the false alarm problem and the problem of being vulnerable to image processing attacks in existing watermark systems will have positive and important significance for promoting the progress of image copyright protection research. Summary of the invention

[0006] The purpose of the present invention is to provide an image blind watermarking method based on an exponential function in order to solve the above problems, which can realize blind watermark extraction while effectively avoiding the watermark false positive problem and effectively resist image processing attacks.

[0007] The present invention achieves the above-mentioned purpose through the following technical solutions, and the specific steps are detailed as follows:

[0008] (1) Watermark embedding process:

[0009] Step 1: Decompose the color host image into sub-blocks of different frequencies through one-dimensional Haar discrete wavelet transform, and obtain the low-frequency components LL in the horizontal and vertical directions, the low-frequency components LH in the horizontal direction and the high-frequency components LH in the vertical direction, the high-frequency components HL in the horizontal direction and the low-frequency components HH in the horizontal and vertical directions.

[0010] [LL,[LH r ,LH g ,LH b ],HL,HH]=DWT(IH,'haar')

[0011] Step 2: The red component LH of the obtained LH r Perform SVD to get (U, S, V).

[0012] [U,S,V]=SVD(LH r )

[0013] Step 3: The original watermark image w is embedded into the singular value matrix S of the host image through the exponential function algorithm to obtain the matrix S' after embedding the watermark.

[0014] S'=(α·w)·e i·S

[0015] Where e represents the exponent, i represents the imaginary number, and α is the watermark embedding strength.

[0016] Step 4: Decompose S' by inverse singular value to obtain the red component LH containing the watermark information r '.

[0017] LH r '=U·S'·V T

[0018] Step 5: Obtain the final host image IW after embedding the watermark through inverse wavelet transform.

[0019] IW=IDWT(LL,[LH r ',LH g ,LH b],HL,HH,'haar')

[0020] (2) Watermark extraction process:

[0021] The first step: decompose the host image embedded with watermark into sub-blocks of different frequencies through one-dimensional Haar discrete wavelet transform.

[0022] [LLT,[LHT r ,LHT g ,LHT b ],HLT,HHT]=DWT(IW,'haar')

[0023] Step 2: The red component of the obtained LHT r Perform singular value decomposition.

[0024] [UT,ST,VT]=SVD(LHT r )

[0025] Step 3: By performing an absolute value operation on the obtained singular value matrix ST, the watermark F can be restored.

[0026] F=|ST| / α

[0027] Where |·| is an absolute value operation. Here α is the watermark embedding strength, which is set to 0.1 in this project. It is worth noting that the method of the present invention directly embeds the watermark to avoid the false positive problem of the watermark. In addition, the watermark extraction process of the image blind watermarking method based on the exponential function of the present invention does not require the participation of the original watermark and the original host, so the method of this project realizes the blind extraction of the watermark. In addition, the proposed image blind watermarking method based on the exponential function can resist image processing attacks, and relevant experimental proofs will be studied later.

[0028] The characteristics and beneficial effects of the present invention are:

[0029] Compared with the proposed additive watermark embedding algorithm, the present invention mainly studies the blind extraction of watermarks without the need for either the original watermark information or the original host information. By proposing a blind watermark module composed of an exponential function embedding algorithm, an image blind watermarking method based on an exponential function is designed. The advantages of the present invention are as follows: (1) Different from the existing watermarking method in which the watermark image is embedded into the host image in an additive manner, this method embeds the watermark based on an exponential function; (2) This watermarking method can realize blind watermark extraction; (3) The method proposed in the present invention can avoid the false positive problem; (4) Compared with the widely used SVD-based watermarking system, the method proposed in the present invention can resist image processing attacks; (5) The exponential function-based embedding algorithm proposed in the present invention can be extended to any additive watermark embedding method. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a schematic diagram of the watermark embedding principle provided by the present invention.

[0031] Figure 2 This is a schematic diagram of the watermark extraction principle provided by the present invention.

[0032] Figure 3 The original watermark image, the original host image, the host image after embedding the watermark, and the watermark image extracted without any attack:

[0033] (a) is the original watermark image;

[0034] (b) is the original host image;

[0035] (c) Host image after embedding watermark;

[0036] (d) The watermarked image extracted without any attack.

[0037] Figure 4 for Figure 3 (c) Watermarked images extracted from the image after being subjected to different image processing attacks.

[0038] (a) is the extraction result when attacked by 0.001 Gaussian noise;

[0039] (b) is the extraction result when attacked by 0.005 Gaussian noise;

[0040] (c) is the extraction result when attacked by 0.01 Gaussian noise;

[0041] (d) is the extraction result when subjected to 6.25% shear attack;

[0042] (e) is the extraction result when subjected to 12.5% ​​shear attack;

[0043] (f) is the extraction result when subjected to 25% shear attack;

[0044] (g) is the extraction result when subjected to sharpening attack with an intensity of 0.6;

[0045] (h) is the extraction result when attacked by low-pass filtering;

[0046] (i) is the extraction result when subjected to 0.5 Gaussian blur attack;

[0047] (j) is the extraction result when attacked by screenshot;

[0048] (k) is the extraction result when subjected to moiré attack.

[0049] In the figure, the components represented by the reference numerals are as follows:

[0050] Figure 1 In: w represents the original watermark image; α represents the watermark embedding strength, which is set to 0.1 in this project; IH represents the original host image; DWT represents discrete wavelet transform; LH represents the low-frequency part of the host image in the horizontal direction and the high-frequency part in the vertical direction; LH r represents the red component of the obtained LH; SVD represents singular value decomposition; (U, S, V) represents LH r The result obtained by singular value decomposition; S' represents the matrix containing watermark information synthesized by exponential function; ISVD represents inverse singular value decomposition; LH r ' represents the red component containing the watermark information obtained by inverse singular value decomposition; IDWT represents inverse discrete wavelet transform; IW represents the host image after the final watermark is embedded obtained by inverse wavelet transform.

[0051] Figure 2 Middle: LHT represents the low frequency in the horizontal direction and the high frequency in the vertical direction of the host image after the watermark is finally embedded; LHT r represents the red component of the obtained LHT; (UT, ST, VT) represents LH r The result obtained by singular value decomposition; F represents the watermark recovered by F = |ST| / α. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] Embodiment 1

[0054] See also Figure 1-2 As shown in Figure 1, a blind image watermarking method based on exponential function is presented. In order to verify the effectiveness and robustness of the method, Figure 3-4 The experimental results of watermark extraction without any attack and after different image processing attacks are given.

[0055] Figure 1The embedding process of the proposed watermarking method mainly includes three stages. The first stage is DWT-SVD. The host image IH is obtained by DWT and SVD to obtain (U, S, V). The second stage is the exponential function embedding process. The original watermark image w is embedded into the singular value matrix S of the host image through the exponential function watermark embedding method, and then the matrix S' containing the watermark information is obtained. The third stage is the inverse process of the first stage. The final host image IW containing the watermark is obtained through inverse singular value decomposition and inverse wavelet transform, completing the embedding of the watermark.

[0056] Figure 2 The extraction process of the proposed watermarking method mainly includes two stages. The first stage is DWT-SVD, where the host image IW containing the watermark is obtained through DWT and SVD (UT, ST, VT). The second stage is the blind extraction process, where the absolute value operation can directly restore the watermark F and complete the blind extraction of the watermark.

[0057] Figure 3 (a) is the original watermark image, Figure 3 (b) is the original host image, Figure 3 (c) is the host image after embedding the watermark. Figure 3 (c) It can be seen that the host image after embedding the watermark is indistinguishable from the original host image to the naked eye. The watermark image extracted without any attack is shown in Figure 2. Figure 3 As shown in (d), it can be seen that when the host image is not attacked in any way, the extracted watermark image and the input watermark image are indistinguishable from each other to the naked eye. In addition, in order to quantitatively evaluate the extraction results, the present invention calculates the correlation coefficient between the original watermark image and the extracted watermark image. By calculation, it is found that the correlation coefficient between the watermark image extracted from the host and the original watermark is 1. The above results show that there is no difference between the extracted watermark image and the original watermark image. The image blind watermarking method based on the exponential function proposed by the present invention can be effectively used for the extraction of color natural host images.

[0058] The attacks adopted by the present invention on the host image after embedding the watermark include: Gaussian noise attacks with intensities of 0.001, 0.005 and 0.01, shear attacks with intensities of 6.25%, 12.5% ​​and 25%, sharpening attacks with an image sharpening intensity of 0.6, Gaussian low-pass filtering attacks with a filtering window of 3×3, Gaussian blur attacks with an intensity of 0.5, screenshot attacks and simulated moiré attacks. Figure 4 for Figure 3(c) The watermark image extracted from the host after being subjected to different image processing attacks. It can be seen that the extracted watermark result still contains the main information of the original watermark. It is worth noting that when the host image embedded with the watermark is attacked by screenshots, the main information of the extracted watermark can still be identified. In order to quantitatively evaluate the extraction results, the present invention calculates the correlation coefficient value between the original watermark image and the extracted watermark image. Through calculation, it is found that the correlation coefficient values ​​between the watermark image extracted from the host and the original watermark are 0.9690, 0.6527, 0.4787, 0.8447, 0.7316, 0.5742, 0.9803, 0.8137, 0.9642, 0.9251, and 0.4788, respectively. The above results show that the image blind watermarking method based on exponential function proposed in the present invention has a certain robustness against various image processing attacks, especially screenshot attacks.

[0059] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0060] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A blind image watermarking method based on an exponential function. This method is mainly aimed at the problems of false positive, non-blind and difficult to resist camouflage attacks in the widely used additive watermark embedding system based on SVD. A new method for fusing watermark and host based on an exponential function is designed. The exponential function embedding algorithm can realize semi-blind extraction of watermark while effectively avoiding the false positive problem of watermark and effectively resisting image processing attacks. To this end, the present invention develops an image blind watermarking system based on an exponential function. The technical solution adopted is that in the watermarking system, an image blind watermarking scheme based on an exponential function is designed. With the help of the characteristics of the exponential function, the watermark image and the host image are used as the base and exponent of the exponential function respectively to obtain the host image after embedding the watermark. In this way, in the extraction process, it is only necessary to take the absolute value operation of the host image after embedding the watermark to restore the watermark information. The entire extraction process avoids the participation of the original watermark and the original host image, and realizes the blind extraction of the watermark. The relevant theorem of the exponential function algorithm adopted by the present invention is as follows: For the square matrix , the matrix can be obtained through the exponential function algorithm . By deduction, we can get: Where e represents the exponent, i represents the imaginary number, and a is a constant greater than 0. Therefore, without the participation of matrix B, the input matrix A can be directly restored through matrix AB. The blind watermarking method based on exponential function proposed in this project uses the above characteristics of the exponential function algorithm to take the watermark image and the host image as matrices A and B. In this way, in the extraction process, only the absolute value operation is needed to restore the watermark information. The entire extraction process avoids the participation of the original watermark and the host image, and realizes the blind extraction of the watermark.

2. According to the method for blind image watermarking based on exponential function in claim 1, the specific watermark embedding process based on exponential function embedding algorithm is described as follows: (1) Watermark embedding process: Step 1: Decompose the color host image into sub-blocks of different frequencies through one-dimensional Haar discrete wavelet transform, and obtain the low-frequency components LL in the horizontal and vertical directions, the low-frequency components LH in the horizontal direction and the high-frequency components LH in the vertical direction, the high-frequency components HL in the horizontal direction and the low-frequency components HH in the horizontal and vertical directions. L L L H r L H g L H b H L H H = D W T I H , ' h a a r ' Step 2: Perform SVD on the red component LH of the obtained LHr to obtain (U, S, V). U S V = S V D L H r Step 3: The original watermark image w is embedded into the singular value matrix S of the host image through the exponential function algorithm to obtain the matrix S' after embedding the watermark. Where e represents the exponent, i represents the imaginary number, and a is the watermark embedding strength. Step 4: Obtain the red component LHr' containing the watermark information by inverse singular value decomposition of S'. Step 5: Obtain the final host image IW after embedding the watermark through inverse wavelet transform. I W = I D W T L L L H r ' L H g L H b H L H H , ' h a a r ' (2) Watermark extraction process: The first step: decompose the host image embedded with watermark into sub-blocks of different frequencies through one-dimensional Haar discrete wavelet transform. L L T L H T r L H T g L H T b H L T H H T = D W T I W , ' h a a r ' Step 2: Perform singular value decomposition on the red component LHTr of the obtained LHT. U T S T V T = S V D L H T r Step 3: By performing an absolute value operation on the obtained singular value matrix ST, the watermark F can be restored. in is an absolute value operation. Here a is the watermark embedding strength, which is set to 0.1 in this project. It is worth noting that the method of the present invention directly embeds the watermark to avoid the false positive problem of the watermark. In addition, the watermark extraction process of the image blind watermarking method based on the exponential function of the present invention does not require the participation of the original watermark and the original host, so the method of this project realizes the blind extraction of the watermark. In addition, the proposed image blind watermarking method based on the exponential function can resist image processing attacks, and relevant experimental proofs will be studied later. The main advantages of the image blind watermarking method based on the exponential function proposed in the present invention are: (1) Different from the existing watermarking method that embeds the watermark image into the host image in an additive embedding manner, this method completes the watermark embedding based on the exponential function; (2) This watermarking method can realize blind watermark extraction; (3) The method proposed in the present invention can avoid the false positive problem; (4) Compared with the widely used SVD-based watermarking system, the method proposed in the present invention can resist image processing attacks; (5) The exponential function-based embedding algorithm proposed in the present invention can be extended to any additive watermark embedding method.

Images