WO2022205853A1 - Medical image digital watermarking method based on perceptual hash and data enhancement - Google Patents

Abstract

A medical image digital watermarking method based on perceptual hash and data enhancement. The medical image digital watermarking method comprises: respectively performing feature extraction of DCT transform and SIFT-DCT transform on an original medical image, and establishing a feature data set of the original medical image by using the perceptual hash; performing Logistic chaotic encryption on an original watermark, embedding watermark information, and obtaining a logical key set; similarly, establishing a feature data set of a medical image to be tested; performing XOR operation on the feature data set and the logical key set of the medical image to be tested, and extracting the encrypted watermark and decrypting the encrypted watermark to obtain a reduced watermark; and performing normalization correlation coefficient calculation on the original watermark and the reduced watermark, and determining the ownership of the original medical image and reading the embedded watermark information. According to the described method, data enhancement, perceptual hash, cryptography and zero watermark technology are combined, better semantic retention is achieved, and better robustness and non-visibility are shown in terms of conventional attacks and geometric attacks.

Description

A Medical Image Digital Watermarking Method Based on Perceptual Hashing and Data Augmentation

This application claims the priority of the Chinese patent application filed on March 31, 2021 with the application number 202110350359.1 and the invention titled "A method for digital watermarking of medical images based on perceptual hashing and data enhancement", all of which are The contents are incorporated herein by reference.

technical field

The invention relates to the field of multimedia signal processing, in particular to a medical image digital watermarking method based on perceptual hashing and data enhancement.

Background technique

With the rapid development of information technology, intelligent medicine and remote diagnosis technologies are becoming more and more mature, and medical diagnosis has begun to be active in the "cloud". Name, gender, age, diagnosis report, medical diagnosis map and other information containing patient privacy) will be disseminated and communicated through the Internet, which will become more convenient and convenient in the future, but the random tampering of information will follow , leakage, copying and other issues, behind these problems are a series of deep concerns about information security, copyright issues, and privacy leakage. In addition, medical image data has unique characteristics. Medical images are an important source of information for medical staff to understand and diagnose the condition of patients. Therefore, the industry has very strict requirements on the quality of medical data used for medical diagnosis. Generally, it is not allowed. Any modification and adjustment of medical data, in addition, the acquisition of medical images often costs a lot, and the cost of a common clinical CT, MRI, and PET examination costs hundreds, or even nearly 1,000 yuan. Therefore, any operation that may cause losses to medical images is not advisable, and the birth of digital zero-watermarking technology just makes up for this gap.

As a new security method, digital zero-watermarking technology can perform security authentication and copyright protection on information products, and has unique advantages for the protection, identification, control and protection of joint interests and responsibilities of today's medical resources. , which can effectively prevent illegal users from intercepting, tampering and illegally copying medical images, which greatly improves the application depth and breadth of medical images in the field of anti-counterfeiting. And zero watermark can better protect medical images from being damaged.

However, some of the current zero-watermarking algorithms are robust to geometric attacks, and some are robust to traditional attacks, but they cannot take into account both attacks at the same time, because of the location and attack characteristics of traditional attacks and geometric attacks. Completely different, there are very few specific zero-watermark algorithms that can resist both types of attacks well at the same time.

SUMMARY OF THE INVENTION

Based on this, it is necessary to provide a medical image digital watermarking method based on perceptual hashing and data enhancement, which can show good robustness and invisibility in both conventional attacks and geometric attacks. The patient's private information is well protected. Its specific plan is as follows:

A medical image digital watermarking method based on perceptual hashing and data enhancement, comprising:

Perform feature extraction of DCT transform and SIFT-DCT transform on the original medical image respectively to obtain a data enhancement feature set of the original medical image;

The data enhancement feature set of the original medical image is correspondingly generated by perceptual hashing to generate the feature binary sequence of the original medical image and the feature data set of the original medical image is established;

Logistic chaotic encryption is performed on the original watermark to obtain an encrypted chaotic scrambled watermark, and the feature data set of the original medical image and the chaotic scrambled watermark are subjected to bit-by-bit XOR operation to obtain a logical key set and embed the watermark information;

Perform feature extraction of DCT transform and SIFT-DCT transform on the medical image to be tested, respectively, to obtain a data enhancement feature set of the medical image to be tested;

The data enhancement feature set of the medical image to be tested is used to generate a binary sequence of features of the medical image to be tested correspondingly by using perceptual hashing, and the feature data set of the medical image to be tested is established;

Performing an XOR operation on the feature data set of the medical image to be tested and the logical key set, extracting an encrypted watermark, and decrypting the extracted encrypted watermark to obtain a restored watermark;

Normalized correlation coefficient calculation is performed on the original watermark and the restored watermark, the ownership of the original medical image is determined, and the embedded watermark information is read.

Preferably, in the above-mentioned medical image digital watermarking method provided in the embodiment of the present invention, the feature extraction of DCT transform and SIFT-DCT transform is performed on the original medical image respectively, and the data enhancement feature set of the original medical image is obtained, which specifically includes:

Extracting the features of the original medical image by DCT transformation to obtain the first data enhancement feature sequence;

The original medical image is subjected to SIFT-DCT transform feature extraction to obtain a second data enhancement feature sequence; the first data enhancement feature sequence and the second data enhancement feature sequence constitute the data enhancement feature of the original medical image. set.

Preferably, in the above-mentioned medical image digital watermarking method provided by the embodiment of the present invention, the data enhancement feature set of the original medical image is used to generate the feature binary sequence of the original medical image correspondingly by using perceptual hashing, and the Feature datasets of raw medical images, including:

Select the upper left corner low frequency coefficient module in the first data enhancement feature sequence, and use the perceptual hash to generate the first feature binary sequence correspondingly;

Select the upper left low frequency coefficient module in the second data enhancement feature sequence, and use perceptual hash to generate a second feature binary sequence correspondingly; the first feature binary sequence and the second feature binary sequence form the original A dataset of features for medical images.

Preferably, in the above-mentioned medical image digital watermarking method provided in the embodiment of the present invention, Logistic chaotic encryption is performed on the original watermark to obtain an encrypted chaotic scrambled watermark, which specifically includes:

Generate chaotic sequences through Logistic Map;

The generated chaotic sequence utilizes a hash function to generate a binary sequence;

XOR scrambling is performed on the pixel position space in the original watermark according to the order of the binary sequence to obtain an encrypted chaotic scrambling watermark.

Preferably, in the above-mentioned medical image digital watermarking method provided in the embodiment of the present invention, a bit-by-bit XOR operation is performed on the feature data set of the original medical image and the chaotic scrambled watermark to obtain a logical key set, which specifically includes: :

Perform bit-by-bit XOR operation on the first characteristic binary sequence and the chaotic scrambled watermark to obtain a first key;

A bit-by-bit XOR operation is performed on the second characteristic binary sequence and the chaotic scrambled watermark to obtain a second key; the first key and the second key form a logical key set.

Preferably, in the above-mentioned medical image digital watermarking method provided in the embodiment of the present invention, after acquiring the logical key set, the method further includes:

The logical key set is used as a key to apply to a third party, and the watermark information is registered in the IPR information base.

Preferably, in the above-mentioned medical image digital watermarking method provided by the embodiment of the present invention, the feature extraction of DCT transform and SIFT-DCT transform is performed on the medical image to be tested, and a data enhancement feature set of the medical image to be tested is obtained, which specifically includes: :

Perform feature extraction of DCT transformation on the medical image to be tested to obtain a third data enhancement feature sequence;

Perform feature extraction of SIFT-DCT transformation on the medical image to be tested to obtain a fourth data enhancement feature sequence; the third data enhancement feature sequence and the fourth data enhancement feature sequence form the data of the medical image to be tested Enhanced feature set.

Preferably, in the above-mentioned medical image digital watermarking method provided in the embodiment of the present invention, the data enhancement feature set of the medical image to be tested is used to generate a binary sequence of features of the medical image to be tested correspondingly by using perceptual hashing, and to establish The feature dataset of the medical image to be tested specifically includes:

Select the upper left corner low frequency coefficient module in the third data enhancement feature sequence, and use perceptual hash to generate the third feature binary sequence correspondingly;

Select the upper left low frequency coefficient module in the fourth data enhancement feature sequence, and use perceptual hash to generate a fourth feature binary sequence correspondingly; the third feature binary sequence and the fourth feature binary sequence form the to-be-to-be A dataset of features for measuring medical images.

Preferably, in the above-mentioned medical image digital watermarking method provided in the embodiment of the present invention, an XOR operation is performed on the feature data set of the medical image to be tested and the logical key set to extract an encrypted watermark, which specifically includes:

XOR operation is performed on the third characteristic binary sequence and the first key to extract the first encrypted watermark;

XOR operation is performed on the fourth characteristic binary sequence and the second key to extract the second encrypted watermark;

After the first encrypted watermark and the second encrypted watermark are discriminated by a comparison discriminator, an encrypted watermark is obtained.

Preferably, in the above-mentioned medical image digital watermarking method provided in the embodiment of the present invention, the extracted encrypted watermark is decrypted to obtain a restored watermark, which specifically includes:

Generate the chaotic sequence through Logistic Map;

using the chaotic sequence to generate the binary sequence using a hash function;

The extracted encrypted watermark is XOR bit by bit according to the sequence of the binary sequence to obtain a restored watermark.

It can be seen from the above technical solutions that a method for digital watermarking of medical images based on perceptual hashing and data enhancement provided by the present invention includes: extracting features of DCT transform and SIFT-DCT transform on the original medical image, respectively, to obtain the original medical image. The data enhancement feature set of medical images; the data enhancement feature set of the original medical image is correspondingly generated by the perceptual hash to generate the feature binary sequence of the original medical image, and the feature data set of the original medical image is established; the original watermark is encrypted by Logistic chaos, The encrypted chaotic scrambled watermark is obtained, and the feature data set of the original medical image and the chaotic scrambled watermark are subjected to bit-by-bit XOR operation to obtain the logical key set and embed the watermark information; the medical image to be tested is subjected to DCT transformation and SIFT respectively. - Feature extraction of DCT transform to obtain the data enhancement feature set of the medical image to be tested; use the perceptual hash to generate the feature binary sequence of the medical image to be tested corresponding to the data enhanced feature set of the medical image to be tested and establish the medical image to be tested The feature data set of the test image; perform XOR operation on the feature data set of the medical image to be tested and the logical key set, extract the encrypted watermark, and decrypt the extracted encrypted watermark to obtain the restored watermark; the original watermark and the restored watermark are normalized Normalized correlation coefficient calculation to determine the ownership of the original medical image and read the embedded watermark information.

The above method provided by the present invention has lower time complexity by combining data enhancement, perceptual hashing, cryptography and zero-watermarking technology, and the obtained codeword has higher efficiency, better semantic retention, and can effectively compensate for The traditional digital watermarking method cannot take into account the shortcomings of anti-traditional attacks and geometric attacks, and the above methods show good robustness and invisibility in both conventional and geometric attacks, and can well protect the patient's private information. , and the original quality of medical images is guaranteed.

The above description is only an overview of the technical solutions of the present invention, in order to be able to understand the technical means of the present invention more clearly, it can be implemented according to the content of the description, and in order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand , the following specific embodiments of the present invention are given.

Description of drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are for the purpose of illustrating preferred embodiments only and are not to be considered limiting of the invention. Also, the same components are denoted by the same reference numerals throughout the drawings. In the attached image:

1 shows a flowchart of a method for digital watermarking of medical images based on perceptual hashing and data enhancement provided by an embodiment of the present invention;

FIG. 2 shows a schematic flowchart of acquiring a feature data set of an original medical image provided by an embodiment of the present invention;

3 shows a schematic flowchart of embedding a watermark and acquiring a logical key set provided by an embodiment of the present invention;

4 shows a schematic flowchart of acquiring a feature data set of a medical image to be tested provided by an embodiment of the present invention;

5 shows a schematic flowchart of extracting and discriminating watermarks provided by an embodiment of the present invention;

FIG. 6 shows an original medical image provided by an embodiment of the present invention;

FIG. 7 shows an original watermark image provided by an embodiment of the present invention;

8 shows an encrypted chaotic scrambled watermark image provided by an embodiment of the present invention;

Fig. 9 shows the watermark extracted without interference provided by an embodiment of the present invention;

FIG. 10 shows a medical image when the Gaussian noise interference intensity is 40% provided by an embodiment of the present invention;

FIG. 11 shows the watermark extracted when the Gaussian noise interference intensity is 40% provided by an embodiment of the present invention;

12 shows a medical image when the compression quality is 1% JPEG compression provided by an embodiment of the present invention;

FIG. 13 shows the watermark extracted during JPEG compression with a compression quality of 1% provided by an embodiment of the present invention;

14 shows a medical image after median filtering with a window size of [5×5] and 20 filtering times provided by an embodiment of the present invention;

15 shows the watermark extracted after median filtering with a window size of [5×5] and 20 filtering times provided by an embodiment of the present invention;

Fig. 16 shows that the window size provided by the embodiment of the present invention is [7x7], and the medical image after the median filter of the number of filtering times is 20;

17 shows the watermark extracted after median filtering with a window size of [7×7] and 20 filtering times provided by an embodiment of the present invention;

FIG. 18 shows a medical image rotated 10° clockwise according to an embodiment of the present invention;

Fig. 19 shows the watermark extracted when rotated 10° clockwise according to an embodiment of the present invention;

FIG. 20 shows a medical image rotated 32° clockwise according to an embodiment of the present invention;

Figure 21 shows the watermark extracted when rotated 32° clockwise according to an embodiment of the present invention;

FIG. 22 shows a medical image zoomed by 0.2 times provided by an embodiment of the present invention;

Fig. 23 shows the watermark extracted when zoomed by 0.2 times provided by an embodiment of the present invention;

FIG. 24 shows a medical image horizontally shifted to the left by 25% provided by an embodiment of the present invention;

Figure 25 shows the watermark extracted when the horizontal left shift is 25% provided by an embodiment of the present invention;

FIG. 26 shows a medical image vertically shifted by 25% provided by an embodiment of the present invention;

FIG. 27 shows the watermark extracted when vertically moving down by 25% provided by an embodiment of the present invention;

FIG. 28 shows a medical image cropped by 20% along the X-axis provided by an embodiment of the present invention;

FIG. 29 shows the watermark extracted when 20% is cut along the X-axis provided by an embodiment of the present invention;

FIG. 30 shows a medical image with a 20% cut along the Y axis provided by an embodiment of the present invention;

FIG. 31 shows the watermark extracted when 20% is cut along the Y-axis according to an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

The present invention provides a medical image digital watermarking method based on perceptual hashing and data enhancement, as shown in Figure 1, comprising the following steps:

S101. Perform feature extraction of DCT transform and SIFT-DCT transform on the original medical image, respectively, to obtain a data enhancement feature set of the original medical image.

In practical applications, before step S101 is performed, the original watermark embedded in the medical image is a meaningful binary text image, denoted as W={W(i,j)|W(i,j)=0,1; 1≤i≤M ₁ , 1≤j≤M ₂ }, M ₁ and M ₂ are the size, length and width of the original watermark image respectively. The watermark is to protect the patient's personal information, which can be hidden in the patient's medical information. image, so as to achieve secure transmission over the network. Meanwhile, in this embodiment, a 512*512 medical image can be selected as the original medical image, denoted as Img(i,j); W(i,j) and Img(i,j) represent the original watermark and the original medical image respectively The pixel gray value of the image.

S102 , using perceptual hashing to generate a feature binary sequence of the original medical image corresponding to the data enhancement feature set of the original medical image, and establish a feature data set of the original medical image.

It should be noted that the perceptual hash technology takes multimedia perception characteristics as invariants, conforms to the human perception model, and more accurately realizes the analysis and protection of multimedia perception content. Steps S101 and S102 are to perform feature vector extraction based on perceptual hashing of data-enhanced DCT and SIFT-DCT on the original medical image Img(i,j) in the transform domain to obtain a feature dataset of the original medical image. The robustness of the watermark can be greatly improved by expanding the feature data set in the form of data enhancement, and by using perceptual hashing, the hash value obtained by perceptual hashing can be used as auxiliary information for embedding and extracting the watermark, and the The watermark is generated by relying on the hash value of the image content as a key. The combination of the two greatly improves the robustness of the watermark against both geometric attacks and traditional attacks.

S103 , perform Logistic chaotic encryption on the original watermark to obtain an encrypted chaotic scrambled watermark, and perform a bit-by-bit XOR operation on the feature data set of the original medical image and the chaotic scrambled watermark to obtain a logical key set and embed the watermark information.

S104. Perform feature extraction of DCT transform and SIFT-DCT transform on the medical image to be tested, respectively, to obtain a data enhancement feature set of the medical image to be tested.

It can be understood that the medical image to be tested here can be considered as a medical image formed after the original medical image may be subject to Gaussian noise interference, median filtering, compression, rotation, translation and other geometric attacks or conventional attacks in the process of network transmission;

S105 , using perceptual hashing to enhance the data feature set of the medical image to be tested, respectively, to generate a binary sequence of features of the medical image to be tested, and to establish a feature data set of the medical image to be tested.

S106 , performing an XOR operation on the feature data set of the medical image to be tested and the logical key set, extracting an encrypted watermark, and decrypting the extracted encrypted watermark to obtain a restored watermark.

S107, perform normalized correlation coefficient calculation on the original watermark and the restored watermark, determine the ownership of the original medical image and read the embedded watermark information.

In the above-mentioned medical image digital watermarking method provided by the embodiment of the present invention, by combining data enhancement, perceptual hashing, cryptography and zero watermarking technology, the time complexity is low, and the obtained codeword has high efficiency, The semantic preservation is good, which effectively makes up for the shortcomings of traditional digital watermarking methods that cannot take into account traditional attacks and geometric attacks, and the method shows good robustness and invisibility in both conventional and geometric attacks. It can well protect the patient's private information and ensure the original quality of medical images.

In this embodiment, the data enhancement method is first used to establish a feature vector data set for the whole image through DCT transformation and SIFT-DCT transformation; then, the information to be embedded is encrypted by Logistic chaos and combined with Hash function, "third-party concept" and image visual feature vector data set, construct zero watermark, and realize the anti-geometric, conventional attack and encryption processing of the constructed digital zero watermark image. The above-mentioned medical image digital watermarking method provided by the embodiment of the present invention includes the construction of the original medical image feature vector set based on the perceptual hash of data-enhanced DCT and SIFT-DCT, watermark chaotic scrambling encryption, watermark embedding, watermark extraction and watermark decryption And restore the five major parts.

In specific implementation, in the above-mentioned medical image digital watermarking method provided by the embodiment of the present invention, step S101 performs DCT (discrete cosine transform) and SIFT-DCT (scale invariant feature transform) on the original medical image Img(i,j) respectively and discrete cosine transform) feature extraction to obtain the data enhancement feature set of the original medical image Img(i,j), as shown in Figure 2, which may specifically include: First, DCT transform the original medical image Img(i,j) to obtain the first data enhancement feature sequence V _DCT (i,j)=DCT(Img(i,j)); then, the original medical image Img(i,j) is subjected to the feature extraction of SIFT-DCT transformation, Obtain the second data enhancement feature sequence V _Sift-DCT (i,j)=SIFI-DCT(Img(i,j)); the first data enhancement feature sequence V _DCT (i,j) and the second data enhancement feature sequence V _Sift-DCT (i,j) composes the data-augmented feature set of the original medical image Img(i,j).

It should be noted that the SIFT (sale invariant feature transform, scale invariant feature transform) operator has ideal robustness to rotation, translation, scaling and projection transformation, and is a description used in the field of image processing. It has scale invariance and can detect key points in the image. It is a local feature description operator. The operator realizes feature matching mainly in the following three processes: First, extract key points: The key points are some very prominent Points that will not disappear due to factors such as illumination, scale, rotation, etc., such as corner points, edge points, bright spots in dark areas, and dark spots in bright areas; this step is to search for image positions on all scale spaces; Identify potential interest points with scale and rotation invariant; second, locate key points and determine feature orientation: at each candidate position, determine the position and scale through a fine-fit model; the selection of key points is based on each keypoint position is assigned one or more directions based on the local gradient direction of the image; all subsequent operations on the image data are transformed relative to the direction, scale, and position of the keypoint, so that Provide invariance to these transformations; thirdly, through the feature vector of each key point, make a pairwise comparison to find several pairs of feature points that match each other, and establish the corresponding relationship between the scenes.

In addition, it should be noted that the working principle of DCT (Discrete Cosine Transform) is to divide the image into parts with different frequencies, including low-frequency, high-frequency and intermediate-frequency coefficients. The discrete cosine transform is an orthogonal transform based on real numbers. The computational complexity of the DCT domain is small, and it has a strong "energy concentration" feature: the energy of most natural signals (including sound and images) is concentrated in the low-frequency part after discrete cosine transformation, which is easy to extract visual feature vectors, and is consistent with Compatible with internationally popular data compression standards (JPEG, MPEG, H261/263), it is easy to implement in the compression domain. The two-dimensional discrete cosine transform (2D-DCT) formula is as follows:

Among them, x, y are spatial sampling frequency domain; u, v are frequency domain sampling values, they are usually represented by pixel square matrix in digital image processing, namely M=N.

Further, in specific implementation, in the above-mentioned medical image digital watermarking method provided by the embodiment of the present invention, after constructing different types of data enhancement sets through step S101, step S102 converts the original medical image Img(i, j) The data enhancement feature set uses perceptual hashing to respectively generate the feature binary sequence of the original medical image Img(i,j) and establish the feature data set of the original medical image Img(i,j), as shown in Figure 2, which can specifically include : select the upper left corner low frequency coefficient module in the first data enhancement feature sequence V _DCT (i,j), and use the perceptual hash to generate the first feature binary sequence; select the second data enhancement feature sequence V _Sift-DCT (i,j ) in the upper left corner of the low-frequency coefficient module, using the perceptual hash corresponding to generate the second feature binary sequence; the first feature binary sequence and the second feature binary sequence constitute the original medical image Img (i, j) feature dataset Vec ( i,j).

It can be understood that, according to human visual characteristics (HVS), low-IF signals have a greater impact on human vision and represent the main features of medical images. Therefore, when the feature extraction of data enhancement is performed on the image, the data of the low-frequency part can be selected, and the selection of the number of low-intermediate frequency coefficients is also related to the size of the original medical image and the correlation between the medical images for the full-image data enhancement feature transformation. Correlation, the smaller the L value, the greater the correlation. In practical applications, the selected upper left corner of the low frequency is the area with the most concentrated energy, and a 4*8 coefficient matrix can be selected. At this time, the third characteristic binary sequence and the fourth characteristic binary sequence obtained are both 32 bits, that is, selecting The length of L is 32. Of course, the length of L may also be other choices, which are determined according to the actual situation, which is not limited here.

The invention combines zero watermark technology and perceptual hash technology, as a technology of information security, can ensure safe transmission, can also realize information authentication, and has practical application in reality.

In specific implementation, in the above-mentioned medical image digital watermarking method provided by the embodiment of the present invention, step S103 performs Logistic chaotic encryption on the original watermark W(i,j) to obtain an encrypted chaotic scrambled watermark EW(i,j) , which may specifically include: first, according to the initial value x ₀ , generate the chaotic sequence X(j) through the Logistic Map; the initial value of the chaotic coefficient is set to 0.2, the growth parameter is 4, and the number of iterations is 1024; The sequence X(j) uses a hash function to generate a binary sequence; after that, XOR scrambles the pixel position space in the original watermark W(i,j) according to the sequence of the binary sequence, and obtains an encrypted chaotic scrambled watermark EW(i,j).

In specific implementation, in the above-mentioned medical image digital watermarking method provided by the embodiment of the present invention, step S103 combines the feature data set Vec(i,j) of the original medical image Img(i,j) and the chaotic scrambled watermark EW(i ,j) perform bit-by-bit XOR operation to obtain the logical key set Key(i,j), as shown in Figure 3, which may specifically include: first, the first characteristic binary sequence and the chaotic scrambled watermark EW(i,j) j) Perform bit-by-bit XOR operation to obtain the first key Key1(i,j); then, perform bit-by-bit XOR operation on the second characteristic binary sequence and the chaotic scrambled watermark EW(i,j) to obtain the first key Two keys Key2(i,j); the first key and the second key form a logical key set Key(i,j):

Key(i,j)=Key1(i,j)∪Key2(i,j);

Key(i,j)=EW(i,j)⊕Vec(i,j).

The feature data set Vec(i,j) and the chaotic scrambled watermark EW(i,j) are XORed bit by bit, the watermark can be embedded in the medical image, that is, the zero watermark, and the logical key can be obtained and saved at the same time Set Key(i,j), and Key(i,j) will be used when extracting the watermark later. After step S103 is performed to obtain the logical key set, the Key(i,j) can be used as a key to apply to a third party, and the watermark information can be registered in the IPR information database, and then the original medical image Img(i,j) can be obtained. Ownership and use rights, so as to achieve the purpose of protecting medical images.

It can be understood that the present invention performs Logistic chaotic encryption on the watermark information to be embedded and combines the hash function, the "third-party concept" and the image visual feature vector to construct a zero watermark, which can realize the anti-resistance of the constructed digital zero watermark image. Geometry, conventional attacks, and cryptographic processing.

Similarly, in the specific implementation, in the above-mentioned medical image digital watermarking method provided by the embodiment of the present invention, step S104 performs DCT (discrete cosine transform) and SIFT-DCT (scale of the medical image Img'(i,j) to be tested respectively. Invariant feature transform and discrete cosine transform) feature extraction to obtain the data enhancement feature set of the medical image to be tested Img' (i, j), as shown in Figure 4, which may specifically include: First, the medical image to be tested Img' (i,j) Perform feature extraction of DCT transformation to obtain a third data enhancement feature sequence V' _DCT (i, j)=DCT(Img(i, j)); j) Perform feature extraction of SIFT-DCT transformation to obtain a fourth data enhancement feature sequence _V'Sift-DCT (i,j)=SIFT-DCT(Img(i,j)); the third data enhancement feature sequence _V'DCT (i,j) and the fourth data-enhanced feature sequence _V'Sift-DCT (i,j) form a data-enhanced feature set of the medical image Img'(i,j) to be tested.

In specific implementation, in the above-mentioned medical image digital watermarking method provided in the embodiment of the present invention, step S105 uses perceptual hashing to generate the data enhancement feature set of the medical image to be tested Img'(i,j) correspondingly to generate the medical image to be tested. The feature binary sequence of Img'(i,j) and the feature data set of the medical image Img'(i,j) to be tested are established, as shown in Figure 4, which may specifically include: selecting a third data enhancement feature sequence V' _DCT The low-frequency coefficient module in the upper left corner (the area with the most concentrated energy) in (i,j) uses perceptual hashing to generate the third feature binary sequence; select the fourth data enhancement feature sequence _V'Sift-DCT (i,j) in The low-frequency coefficient module in the upper left corner uses the perceptual hash to correspond to generate the fourth feature binary sequence; the third feature binary sequence and the fourth feature binary sequence form the feature dataset Vec' of the medical image to be tested Img'(i,j) (i,j). In practical applications, the selected low-frequency region in the upper left corner is the area with the most concentrated energy, and a 4*8 coefficient matrix can be selected. At this time, the obtained third characteristic binary sequence and fourth characteristic binary sequence are both 32 bits, that is, It is said that the low-frequency 4*8 coefficient modules in the upper left corner are respectively selected, and the feature data set Vec'(i,j) of the medical image to be tested can be obtained through perceptual hashing, which realizes the extraction of the visual feature sequence of the medical image to be tested.

In specific implementation, in the above-mentioned medical image digital watermarking method provided by the embodiment of the present invention, step S106 combines the feature data set Vec'(i,j) and the logical key set of the medical image Img'(i,j) to be tested Key(i,j) performs XOR operation to extract the encrypted watermark EW'(i,j), EW'(i,j)=Key(i,j)⊕Vec'(i,j), as shown in Figure 5 Specifically, it may include: first, performing an XOR operation on the third feature binary sequence and the first key Key1(i,j) to extract the first encrypted watermark; then, combining the fourth feature binary sequence with the second The key Key2(i,j) performs XOR operation to extract the second encrypted watermark; finally, after the first encrypted watermark and the second encrypted watermark are discriminated by the comparison discriminator, the encrypted watermark EW'(i,j) is obtained . This method only needs the key Key(i,j) when extracting the watermark, and does not require the participation of the original medical image, and is a zero watermark extraction algorithm. It should be noted that, the comparison discriminator may have various discrimination methods, such as comparison of sharpness, comparison of resolution, etc., which are not limited here.

During specific implementation, in the above-mentioned medical image digital watermarking method provided by the embodiment of the present invention, step S106 decrypts the extracted encrypted watermark to obtain a restored watermark, which may specifically include: first, according to the known initial value x ₀ , the chaotic coefficient The initial value, growth parameters and the number of iterations are consistent with the encryption process, and the same chaotic sequence X(j) is generated by Logistic Map; the chaotic sequence X(j) is generated by the hash function to the same binary sequence; The sequence sequence performs bit-wise exclusive OR on the extracted encrypted watermark EW'(i,j) to obtain the restored watermark W'(i,j).

Specifically, according to the specific description of each step above, it can be simply understood as: first, perform data enhancement on the original medical image Img(i, j), that is, perform feature extraction of DCT transform and SIFT-DCT transform to obtain feature sets, and then use perception Hash, corresponding to generate the characteristic binary sequence of 32-bit medical images and establish a strong robust feature data set Vec(i,j); then use the properties of Logistic Map to scramble and encrypt the watermark in the frequency domain, combined with the zero watermark The method uses the established picture feature set to embed and extract the watermark. The key obtained after the watermark is embedded can be stored in a third party to ensure the copyright of the picture and the security of the watermark.

Next, step S107 can be performed by calculating the normalized correlation coefficient NC of the original watermark W(i,j) and the restored watermark W'(i,j), and the ownership of the original medical image can be determined by the level of the NC coefficient, and read Get the embedded watermark information.

It should be noted that the normalized cross-correlation (NC) method is used to measure the quantitative similarity between the embedded original watermark W(i,j) and the restored watermark W'(i,j), which is defined as :

Among them, W(i,j) represents the feature vector of the original watermark image, and its length is 32bit; W'(i,j) represents the feature vector of the restored watermark image, which is also 32bit. The normalized correlation coefficient is a method to measure the similarity of two images. By calculating the normalized correlation coefficient, the data can be used to objectively evaluate the similarity of the images.

The present invention will be further described below with reference to the accompanying drawings: as shown in FIG. 6 , the experimental test object is a 512*512 abdominal medical image, represented by Img(i,j), where 1≤i, j≤512. Select a meaningful binary image as the original watermark, denoted as: W={W(i,j)|W(i,j)=0,1; 1≤i≤M ₁ , 1≤j≤M ₂ } , as shown in Figure 7, where the size of the watermark is 32*32.

Firstly, DCT transform and SIFT-DCT transform are applied to the whole image by data enhancement method, and then perceptual hashing is used to generate the characteristic binary sequence of 32-bit medical images respectively and establish the dataset Vec(i,j). Next, set the initial value of the chaos coefficient to 0.2, the growth parameter to 4, and the number of iterations to 1024. Then perform Logistic chaotic encryption on the original watermark W(i,j), and the encrypted chaotic scrambled watermark EW(i,j) is shown in Figure 8. After W'(i,j) is detected by the watermark algorithm, the normalized correlation coefficient NC is calculated to determine whether there is a watermark embedded. When the value is closer to 1, the similarity is higher, so as to judge the robustness of the algorithm. . The distortion degree of the picture expressed by PSNR, when the PSNR value is larger, the distortion degree of the picture is smaller.

Figure 9 shows the watermark extracted when no interference is added. It can be seen that NC=1.00, and the watermark can be extracted accurately.

The anti-conventional attack ability and anti-geometric attack ability of the digital watermarking method are judged below through specific examples.

First, add Gaussian noise: use the imnoise() function to add Gaussian noise to the watermark.

Table 1 shows the experimental data of watermarking against Gaussian noise. It can be seen from Table 1 that when the Gaussian noise intensity is as high as 40%, the PSNR of the image after the attack drops to 8.00dB. At this time, the extracted watermark, the correlation coefficient NC=0.87, can still accurately extract the watermark, and the overall data Both are at 0.8. This shows that the invention has good robustness against Gaussian interference. Figure 10 shows the medical image when the Gaussian noise intensity is 40%, which is visually distinct from the original abdominal medical image; Figure 11 shows the extracted watermark when the Gaussian noise intensity is 40%, NC=0.87.

Table 1 Watermark anti-Gaussian noise interference data

Noise intensity (%)	5	10	15	20	25	30	35	40
PSNR(dB)	15.41	12.65	11.04	10.08	9.37	8.78	8.32	8.00
NC	0.95	0.95	0.95	0.95	0.93	0.93	0.80	0.87

Second, JPEG compression processing

Using the image compression quality percentage as a parameter to perform JPEG compression on abdominal medical images; Table 2 is the experimental data of watermark-resistant JPEG compression. When the compression quality is 1%, the image quality is very low, and the watermark can still be extracted clearly, NC=0.93. Fig. 12 shows a medical image with a compression quality of 1%; Fig. 13 shows a watermark extracted with a compression quality of 1%, NC=0.93, the watermark can be extracted accurately.

Table 2 Watermark anti-JPEG compression experimental data

Compression quality (%)	1	2	5	10	15	20	25	30
PSNR(dB)	25.64	25.64	26.91	30.58	31.81	33.03	33.87	34.53
NC	0.93	0.93	1.00	1.00	1.00	1.00	1.00	1.00

Third, median filter processing

Table 3 shows the anti-median filtering capability of watermarking for medical images. It can be seen from Table 3 that the median filtering parameter is [5x5] and the number of filter repetitions is 20, the existence of the watermark can still be measured, NC=1.00. Figure 14 shows a medical image with the median filter parameter [5x5] and the number of filter repetitions is 20, and the image is still very clear; Figure 15 shows that the median filter parameter is [5x5] and the number of filter repetitions is 20. Watermark, NC=1.00, can extract the watermark. Figure 16 shows a medical image with the median filter parameter [7x7] and the number of filter repetitions is 20; Figure 17 shows the watermark extracted when the median filter parameter is [7x7] and the number of filter repetitions is 20, NC=0.93 , you can extract the watermark.

Table 3 Experimental data of watermark anti-median filtering

Fourth, the rotation transformation

Table 4 shows the experimental data of watermarking against rotation attack. It can be seen from Table 4 that when the image is rotated 32° clockwise, NC=0.88, and the watermark can still be extracted well. Fig. 18 shows a medical image rotated clockwise by 10°; Fig. 19 shows a watermark extracted by a clockwise rotation of 10°, NC=1.00, the watermark can be extracted very clearly. Figure 20 shows a medical image rotated 32° clockwise; Figure 21 shows the watermark extracted by 32° clockwise rotation, NC=0.88, the watermark can be extracted accurately.

Table 4 Watermarking anti-rotation attack experimental data

Degree of rotation°	2°	5°	10°	15°	20°	25°
PSNR(dB)	21.56	18.45	16.68	15.86	15.34	15.04
NC	1.00	1.00	1.00	0.95	0.95	0.88

Fifth, scaling transformation

Table 5 is the experimental data of anti-scaling attack of watermark in medical images. It can be seen from Table 5 that when the scaling factor is as small as 0.2, the correlation coefficient NC=1.00, and the watermark can be extracted well. Figure 22 shows the scaled medical image (the scaling factor is 0.2); Figure 23 shows the watermark extracted after the scaling attack, NC=1.00, the watermark can be extracted very accurately.

Table 5 Experimental data of anti-scaling attack of watermark

scaling factor	0.2	0.4	0.6	0.8	1.2	1.4	1.6	1.8	2.2	3.0
NC	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00

Sixth, translation transformation

Table 6 is the experimental data of watermark anti-translation transformation. It can be known from Table 6 that when the image data is moved vertically by 25%, the NC value is higher than 0.70, and the watermark can be extracted accurately, so the watermark method has strong anti-translational transformation ability. Figure 24 shows the image after the medical image is horizontally shifted to the left by 25%; Figure 25 shows the extracted watermark after the horizontal shift to the left by 25%, the watermark can be extracted accurately, NC=0.95. Fig. 26 shows the image after the medical image is vertically moved down by 25%; Fig. 27 shows the extracted watermark after the vertical move down by 25%, the watermark can be extracted accurately, NC=0.88.

Table 6 Experimental data of watermark anti-translation transformation

Seventh, cut attack

Table 7 is the experimental data of watermark anti-shearing attack. It can be seen from Table 7 that when the image is cut in squares, when the shearing degree along the X axis reaches 20%, the NC value can still reach 0.61. When the shaft shearing degree reaches 20%, the NC value can reach 0.77, and the extracted watermark can still be distinguished, indicating that the watermarking algorithm has a strong ability of shearing attack. Figure 28 shows the medical image after cutting 20% of the image along the X-axis; Figure 29 shows the watermark extracted after cutting the 20% image along the X-axis, the watermark can be extracted accurately, NC=0.61. Figure 30 shows the medical image after cutting 20% of the image along the Y axis; Figure 31 shows the watermark extracted after cutting the 20% image along the Y axis, the watermark can be extracted accurately, NC=0.77.

Table 7 Experimental data of watermarking against shearing attack in Y-axis direction

It can be seen from the above description that the present invention is based on the perceptual hash and data enhancement medical image digital watermarking technology, enhances the feature set of the extracted feature sequence, has good robustness, and is suitable for Gaussian noise interference, JPEG compression processing, medium Conventional attacks such as value filtering, as well as geometric attacks such as rotation transformation, scaling transformation, translation transformation, and shearing attack, can still accurately extract watermarks, and have strong resistance to conventional attacks and geometric attacks.

Professionals may further realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two, in order to clearly illustrate the possibilities of hardware and software. Interchangeability, the above description has generally described the components and steps of each example in terms of functionality. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of this application.

The steps of a method or algorithm described in conjunction with the embodiments disclosed herein may be directly implemented in hardware, a software module executed by a processor, or a combination of the two. A software module can be placed in random access memory (RAM), internal memory, read only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other in the technical field. in any other known form of storage medium.

A method for digital watermarking of medical images based on perceptual hashing and data enhancement provided by an embodiment of the present invention includes: extracting features of DCT transform and SIFT-DCT transform on an original medical image, respectively, to obtain a data enhancement feature set of the original medical image ; The data enhancement feature set of the original medical image is correspondingly generated by the perceptual hash to generate the feature binary sequence of the original medical image, and the feature data set of the original medical image is established; Logistic chaotic encryption is performed on the original watermark, and the encrypted chaotic scrambling is obtained. Watermark, and perform bit-by-bit XOR operation on the feature data set of the original medical image and the chaotic scrambled watermark to obtain the logical key set and embed the watermark information; perform DCT transform and SIFT-DCT transform feature extraction on the medical image to be tested, respectively, Obtain the data enhancement feature set of the medical image to be tested; use the perceptual hash to generate the feature binary sequence of the medical image to be tested correspondingly to the data enhanced feature set of the medical image to be tested, and establish the feature data set of the medical image to be tested; Perform XOR operation on the feature data set and logical key set of the measured medical image, extract the encrypted watermark, decrypt the extracted encrypted watermark, and obtain the restored watermark; calculate the normalized correlation coefficient of the original watermark and the restored watermark to determine Ownership of original medical images and read embedded watermark information. By combining data enhancement, perceptual hashing, cryptography and zero-watermarking technology, the method has lower time complexity, and the obtained codewords have higher efficiency and better semantic retention, effectively making up for the traditional digital The watermarking method cannot take into account the shortcomings of anti-traditional attacks and geometric attacks, and the method shows good robustness and invisibility in both conventional and geometric attacks, which can well protect the patient's private information and ensure medical The original quality of the image.

Finally, it should also be noted that in this document, relational terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply these entities or that there is any such actual relationship or sequence between operations. Moreover, the terms "comprising", "comprising" or any other variation thereof are intended to encompass a non-exclusive inclusion such that a process, method, article or device that includes a list of elements includes not only those elements, but also includes not explicitly listed or other elements inherent to such a process, method, article or apparatus. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in a process, method, article or apparatus that includes the element.

The above-mentioned embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it is still possible to implement the foregoing implementations. The technical solutions described in the examples are modified, or some technical features thereof are equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the within the protection scope of the present invention.

Claims (10)

1. A method for digital watermarking of medical images based on perceptual hashing and data enhancement, comprising:

   Perform feature extraction of DCT transform and SIFT-DCT transform on the original medical image respectively to obtain a data enhancement feature set of the original medical image;

   The data enhancement feature set of the original medical image is correspondingly generated by perceptual hashing to generate the feature binary sequence of the original medical image and the feature data set of the original medical image is established;

   Logistic chaotic encryption is performed on the original watermark to obtain an encrypted chaotic scrambled watermark, and the feature data set of the original medical image and the chaotic scrambled watermark are subjected to bit-by-bit XOR operation to obtain a logical key set and embed the watermark information;

   Perform feature extraction of DCT transform and SIFT-DCT transform on the medical image to be tested, respectively, to obtain a data enhancement feature set of the medical image to be tested;

   The data enhancement feature set of the medical image to be tested is used to generate a binary sequence of features of the medical image to be tested correspondingly by using perceptual hashing, and the feature data set of the medical image to be tested is established;

   Performing an XOR operation on the feature data set of the medical image to be tested and the logical key set, extracting an encrypted watermark, and decrypting the extracted encrypted watermark to obtain a restored watermark;

   Normalized correlation coefficient calculation is performed on the original watermark and the restored watermark, the ownership of the original medical image is determined, and the embedded watermark information is read.
2. The method for digital watermarking of medical images according to claim 1, wherein the feature extraction of DCT transform and SIFT-DCT transform is performed on the original medical image respectively to obtain a data enhancement feature set of the original medical image, which specifically includes:

   Extracting the features of the original medical image by DCT transformation to obtain the first data enhancement feature sequence;

   The original medical image is subjected to SIFT-DCT transform feature extraction to obtain a second data enhancement feature sequence; the first data enhancement feature sequence and the second data enhancement feature sequence constitute the data enhancement feature of the original medical image. set.
3. The method for digital watermarking of medical images according to claim 2, wherein the data-enhanced feature set of the original medical image is correspondingly generated by using perceptual hashing to generate binary feature sequences of the original medical image, and the original medical image is established. Feature datasets of medical images, including:

   Select the upper left corner low frequency coefficient module in the first data enhancement feature sequence, and use the perceptual hash to generate the first feature binary sequence correspondingly;

   Select the upper left low frequency coefficient module in the second data enhancement feature sequence, and use perceptual hash to generate a second feature binary sequence correspondingly; the first feature binary sequence and the second feature binary sequence form the original A dataset of features for medical images.
4. The medical image digital watermarking method according to claim 3, wherein the original watermark is subjected to Logistic chaotic encryption to obtain an encrypted chaotic scrambled watermark, which specifically comprises:

   Generate chaotic sequences through Logistic Map;

   The generated chaotic sequence utilizes a hash function to generate a binary sequence;

   XOR scrambling is performed on the pixel position space in the original watermark according to the order of the binary sequence to obtain an encrypted chaotic scrambling watermark.
5. The method for digital watermarking of medical images according to claim 4, wherein the characteristic data set of the original medical image and the chaotic scrambled watermark are subjected to a bit-by-bit XOR operation to obtain a logical key set, which specifically includes:

   Perform bit-by-bit XOR operation on the first characteristic binary sequence and the chaotic scrambled watermark to obtain a first key;

   A bit-by-bit XOR operation is performed on the second characteristic binary sequence and the chaotic scrambled watermark to obtain a second key; the first key and the second key form a logical key set.
6. The medical image digital watermarking method according to claim 5, characterized in that, after acquiring the logical key set, further comprising:

   The logical key set is used as a key to apply to a third party, and the watermark information is registered in the IPR information base.
7. The method for digital watermarking of medical images according to claim 6, wherein the feature extraction of DCT transform and SIFT-DCT transform is performed on the medical image to be tested, respectively, to obtain a data enhancement feature set of the medical image to be tested, specifically comprising:

   Perform feature extraction of DCT transformation on the medical image to be tested to obtain a third data enhancement feature sequence;

   Perform feature extraction of SIFT-DCT transformation on the medical image to be tested to obtain a fourth data enhancement feature sequence; the third data enhancement feature sequence and the fourth data enhancement feature sequence form the data of the medical image to be tested Enhanced feature set.
8. The method for digital watermarking of medical images according to claim 7, wherein the data-enhanced feature set of the medical image to be tested is used to generate a binary sequence of features of the medical image to be tested correspondingly by using perceptual hashing, and the set of features of the medical image to be tested is established. The feature dataset of the medical image to be tested, specifically including:

   Select the upper left corner low frequency coefficient module in the third data enhancement feature sequence, and use perceptual hash to generate the third feature binary sequence correspondingly;

   Select the upper left low frequency coefficient module in the fourth data enhancement feature sequence, and use perceptual hash to generate a fourth feature binary sequence correspondingly; the third feature binary sequence and the fourth feature binary sequence form the to-be-to-be A dataset of features for measuring medical images.
9. The method for digital watermarking of medical images according to claim 8, wherein an XOR operation is performed on the characteristic data set of the medical image to be tested and the logical key set to extract an encrypted watermark, which specifically includes:

   XOR operation is performed on the third characteristic binary sequence and the first key to extract the first encrypted watermark;

   XOR operation is performed on the fourth characteristic binary sequence and the second key to extract the second encrypted watermark;

   After the first encrypted watermark and the second encrypted watermark are discriminated by a comparison discriminator, an encrypted watermark is obtained.
10. The method for digital watermarking of medical images according to claim 9, wherein the extracted encrypted watermark is decrypted to obtain a restored watermark, which specifically comprises:

    Generate the chaotic sequence through Logistic Map;

    using the chaotic sequence to generate the binary sequence using a hash function;

    The extracted encrypted watermark is XOR bit by bit according to the sequence of the binary sequence to obtain a restored watermark.

Images