Crypto-Watermarking Algorithm Using Weber’s Law and AES: A View to Transfer Safe Medical Image

We propose a novel method for medical image watermarking in the DCTdomain using the AES encryption algorithm. First, we decompose the original medical image into subblocks of 8× 8. Besides, we apply the DCT and the quantization, respectively, to each subblock. However, in the DCTdomain, an adequate choice of the DCTcoefficients according to the quantization table in the middle frequencies band is performed. After that, we embed the patient’s data into the corresponding medical image. *e insertion step is carried out just after the quantization phase. To increase the robustness, we encrypt the watermarked medical images by using the AES algorithm based on chaotic technique. Arnold’s cat map is used to shuffle the pixel values, and a chaotic Henonmap is utilized to generate an aleatory sequence for the AES algorithm.*e shuffled watermarked image is encrypted using the modified AES algorithm.*e constant ofWeber is used to choose the suitable visibility factor for embedding a watermark with high robustness. To control identification, after application of attacks, we use the serial turbo code for correction of the watermark to recover the data inserted. *e average peak signal-to-noise ratio (PSNR) of the medical images obtained is 61,7769 dB. Experimental results demonstrate the robustness of the proposed schema against various types of attacks.


Introduction
Recently, the rapid development of communication technologies and multimedia resulted in the transmission of great amount of multimedia data (image, audio, videos, etc.). Sometimes this multimedia information is sensitive in nature such as medical imaging. en, it must be protected against modification and reproduction. erefore, it is essential to find an efficient method able to protect sensitive information from unauthorized use. Typical mechanisms given these security aspects are cryptography [1,2] and watermarking [3,4]. e watermark digital image, which deals with the incorporation of special data (watermark) in a roof deck, is the most viable solution to the problems of intellectual property and content authentication. e main requirements of a watermarking technique are imperceptibility, robustness, and security [5,6].
Digital watermarking techniques are classified in many ways [7][8][9][10] but one of the most important classifications of a watermarking system is based on whether the watermark is embedded directly by manipulating pixels of the cover image or by transforming the cover image into the frequency domain by various transformation tools (FFT, DCT, DWT, and KLT) and then embedding the watermark by modifying various frequency coefficients. e former method of embedding the watermark is referred as spatial domain watermarking, while as the latter is known as transform domain watermarking. Spatial domain embedding is very simple, computationally efficient, and has high embedding capacity [11]. However, the spatial domain-based watermarking techniques are not robust to most of the image processing operations. Commonly used transform domain techniques are discrete cosine transform (DCT), discrete wavelet transform (DWT), and Karhunen Loeve transforms (KLT) [12][13][14].
Conventionally, cryptographic solutions are used to enhance the security of multimedia data [15][16][17]. In cryptography, the message is scrambled so that it cannot be revealed to third parties malicious. An emerging trend for information security focuses on the combination of cryptographic and steganographic approaches [17]. Steganography differs from cryptography in many ways; among them, the data are not disguised in an unreadable form; therefore, they attract less attention from the opponent. Currently, steganography attracts the attention of researchers, which is growing for various applications. In a typical image-based steganography system, the content of interest is seamlessly inserted into the cover images and then transmitted to the destination without compromising its security and robustness.
In this context, a novel schema of the medical image using cryptography and watermarking algorithm in the DCT domain is showed, which uses constant's Weber to locate the suitable value of the visibility factor for embedding the watermark.
e watermark is integrated in the middle frequencies' band. Cryptography has been projected as a complementary mechanism to improve the security of medical images. Experiments and analyses results show that the watermarking method is robust to several attacks.
is manuscript is organized as follows. Section 1 presents the previous works related to watermarking and encryption algorithm. Section 2 is focused to describe the DCT domain, the AES encryption, and the constant of Weber. In Section 3, we detail the proposed approach. Section 4 shows the experiments results of our method and analyses the robustness of this technique against different attacks in Section 5. And, finally, we conclude our paper.

Literature Survey
In literature, diverse methods have been tried out for image digital watermarking based on discrete cosine transform. In [18], the authors presented a reversible watermarking algorithm based on integer DCT (IntDCT) for medical image protection. In this paper, the first step consists of dividing the original medical image into nonoverlapping blocks. After that, apply the IntDCT on each block. Calculate the energy of each block, select the blocks adequate with the specific threshold, and create the index of adequate blocks. e watermark image is transformed into the binary sequence by using zigzag scan. Finally, embed the secret message into the selected blocks.
In [19], Shang and Kang have proposed a robust watermarking system for medical image security based on DCT and Arnold transformation. Firstly, this method encrypts the watermark data by using the scrambling technology as pretreatment. ereafter, apply the DCT transform on the original image. Extract a visual feature vector through graphic DCT transformation. en, the visual feature vector is combined with the encrypted watermark to generate a binary series by using the hash function. e binary sequence is regarded as the key to exact correctly the watermark.
In [20], the authors have proposed a novel approach of the biomedical watermarking method which combines cryptography with frequency domain watermarking. is algorithm permits to embed EPR (electronic patient records) in the original image by using DCT watermarking and RSA encryption algorithm. e first step consists of converting the EPR to cipher text through the RSA algorithm and then transformed to a binary message to form the watermark. After that, the watermark is converted into its corresponding frequency domain component by using DCT. is information is saved at the higher portion of the original image along with the ciphered keys. On the contrary, the ROI (region of interest) of the infected region is detected by using a contour detection algorithm and combination of region growing. Finally, this binary image (obtained ROI) is transformed using DCT to create another watermark data, and it is saved at the lower part of the original image.
Chittaranjan et al. [21] have proposed a double encryption system in the DCTdomain. is algorithm has been used crosschaos map and Arnold cat map for encrypting the watermark image twice. After that, the encrypted image is embedded into the DCT domain. e same authors have also proposed in [22] a comparative analysis of the watermarking system in DCTand DWT domain using enhanced playfair cipher. e proposed preprocessing stage is implemented in diverse transform domain techniques to embed ownership information in the original media. e encrypted image is formed by applying modified playfair cipher to the watermark image which will be further hidden in cover media using DWT and DCT. e authors in [23] proposed a new blind digital watermarking system based on DCT. is algorithm consists of spreading the original watermark to a number which is more faithful to the image. Next, preprocess the original image by applying the first level of DCT transform to the image and select a portion of transform coefficients. After that, apply the second level of DCT transform to the coefficient image and choose the image feature vector. Finally, embed the spread watermark vector into the image feature vector.
In [24], the authors proposed a blind image watermarking method based on DCT with spread spectrum technique.
is method divides the original image into nonoverlapping blocks. en, each block is changed into the frequency domain by using DCT transform. e medium band of coefficients in the frequency domain of each block is used to embed the watermark.
In [16], the authors proposed an approach based on the new left data mapping (LDM), pixel repetition method (PRM), RC4 encryption, and checksum calculation. e PRM is used to scale the image. Binary secret data is encrypted using the RC4 encryption algorithm; then, the encrypted data is grouped into 3-bit blocks and converted to decimal equivalents. Before integration, these decimal digits are coded by LDM. A fragile logo is integrated into the cover images in addition to the EHR to facilitate early detection of tampering. e paper in [25] presented two different watermarking algorithms for medical images in the transform domain. In first technique, a digital watermark and electronic patient record have been integrated in both regions, region of interest and region of noninterest. In the second technique, the region of interest is kept intact for remote diagnosis purposes and the region of noninterest is used to mask the digital watermark and the EPR. In either algorithm, a discrete cosine transform based on an 8 × 8 block was used.

Scientific Programming
In [26], the authors proposed a medical image watermarking system based on discrete wavelet transform. Before the incorporation of the watermark, a medical image is subjected to a comprehensive DWT. e Haar wavelet was used for this purpose. e LH subband has been divided into 8 × 8 blocks. e coefficients of each 8 × 8 block (LH) are compared and, if necessary, swapped to incorporate the binary digit "1" or "0."

The Fundamental Theory
is part is a fundamental theory for the selected tools: DCT, Weber's constant, AES, Arnold's cat map, and key generation using the chaotic Henon map which are used in this work.

e Discrete Cosine Transform (DCT).
Discrete cosine transform transforms the space domain into its frequency components. e DCT decomposes the image into various frequency bands. e DCT decomposes the image into various frequency bands. Data hiding may be added into low, mid, or high frequency. Frequency bands of 8 × 8 block of DCT transform is illustrated in Figure 1. e secret data is hidden in middle band frequency of DCT coefficient blocks. e formulae for calculating 2D DCT in the watermarking system is defined by equation (1) and inverse 2D DCT transformation is given by equation (2):

Weber's Constant.
Weber's constant has been applied for watermark authentication by Ekta and Suneja [27]. Weber declared that the ratio of augmentation threshold to the intensity is constant [28]. Chen et al. [29] have been using Weber's law to describe Weber's descriptor. It is based on two components: orientation (λ) and differential excitation (χ) of a pixel. To calculate the differential excitation value χ of the pixel (x i , y j ) possessing intensity f(x i , y j ), its difference is computed with intensity of the eight neighbours (shown in Figure 2). en, the ratio of differences relative to the intensity of current pixel (x i , y j ) is assessed. e inverse tangent is next applied on this ratio, as given in the following equation: e value of the differential excitation χ(x i , y j ) is comprised between [−π/2, π/2t]. If the numerator together with denominator in this ratio assesses to zero, in that case, χ(x i , y j ) is taken as zero. e value of the differential excitation χ(x i , y j ) may be negative or positive. Indeed, e value of differential excitation χ(x i , y j ) is positive, if the intensity of the current pixel is lower than the intensities of the neighbours of current pixels.
Diverse other medical images are used in the approved experiments in the watermarking process, with diverse classified zones' numbers as exposed in Figure 3. ese figures illustrate medical images with textures. Figure 3(b) divided into four different zones where the watermark can be embedded with several different visibility factors. Figure 3(d) is partitioned into five zones. In each image, the value of the visibility factor, where the watermarks have to be embedded, changes with different zones.

Arnold's Cat Map.
In the 1960s, Vladimir Arnold discovered two-dimensional invertible chaotic maps. e effect of the map was proved using a cat image, which is why it is called Arnold's cat map [30]. In the matrix form, Arnold's cat map is defined as follows: Scientific Programming where d and c are positive integer constants and n � 0, 1, 2, 3, . . .. For a medical image, (x n , y n ) is an actual pixel value for the medical image at the specified position and (x n+1 , y n+1 ) is the transformed pixel value for the medical image. After execution of N iterations, there are T positive integers such that (x n+1 , y n+1 ) � (x n , y n ). e period T depends on the size of the input medical image and parameters d and c. Arnold's cat map is very efficient for pixel value mixing since it is a simple transformation using a single mode operation. Nevertheless, it has some security weakness, and an attacker can reproduce the original information by iterating the map repeatedly. erefore, the AES algorithm and the chaotic key generator have been applied to realize higher level diffusion properties and confusion in the encryption.

Key Generation Using Chaotic Henon Map. In 1978,
Henon discovered a discrete-time dynamical scheme that exhibited chaotic behavior [31], which is described as It depends on two primordial parameters: λ and a. e pseudorandom behavior of the Henon map has values of λ � 0.3 and 1.07 ≤ a ≤ 1.09. In our work, the concept of exploitation of a key generator founded on a Henon chaotic map is presented to improve the encryption quality of the AES algorithm without changing the actual implementation.
is results in higher confusion and diffusion properties with efficiency and expectable performance. e proposed key generator produces variable length secret keys of diverse sizes (128, 192, and 256 bits), as supported in the AES block encryption algorithm. e single secret key, K s , is defined as Here, r is the size of key K s in bits. A secret key (with a length of 128, 192, or 256 bits) is suitable and long enough. Standard). AES is one of the most known encryption systems for data protection. Invented in 1998 by Joan Daemen and Vincet Rijmen and proved in 2000 by NIST [32][33][34], AES encryption has been widely deployed thanks to its high performance. In effect, AES is reliable which offers high level security, high robustness, wide portability, fast, and easy to implement. Recently, many encryption systems have previously been developed founded on AES for data security.

AES (Advanced Encryption
Different extensions have been established for the AES algorithm by diverse researchers. Nowadays, the AES algorithm is capable of being modified or used to support variable length information and keys. AES tolerates a 128, 192, or 256 bits data length that can be separated into 4 operation blocks organized in 4 × 4 matrix of bytes array (called a state). For encryption and decryption processes, the key and data blocks are passed through N r rounds (N r � 10, 12, or 14), which are extracted from key size K of length (N k � 4, 6, or 8, respectively) in our approach, and N c � 4 characterizes the number of columns in a state matrix depending upon the block size (equal to 32 bits).
As shown in Figure 4, the AES encryption method is defined as numerous iterative transformation rounds (called a round function). Transformation rounds are created on the key size, which is utilized to calculate the number of rounds (N r rounds). e encryption method converts the information blocks into unreadable information blocks. e round function consists of numerous suboperations called    Scientific Programming AddRoundKey, ShiftRows, ByteSub, and MixColumns transformation. e round function is applied after the initial AddRoundKey operation, which depends on the secret key. Correspondingly, the decryption method uses the same order to transform the encrypted information blocks into the actual information blocks using the same secret key.

Proposed Approach
In this section, we explain the proposed algorithm for embedding the totality of the patient's data in a medical image. Before the insertion, the patient's data undergoes many steps with the aim at increasing the level of integration and better ensuring its extraction after applying different attacks. e insertion procedure is done by adding to the image frequency components, proper to the original image data related to the patient. Equation (7) explains the integration procedure of the watermark on the image frequency coefficients: with σ i being the i th old coefficient to support the watermark, σ i ′ being the i th new watermarked coefficient, W i being the i th bit to hide, and α being named often as the visibility factor. Indeed, α is an important factor in the watermarking system. If this factor is big, we win in terms of the robustness; however, we lose in terms of imperceptibility and vice versa. e global proposed schema process is shown in Figure 5.

Preparation of the Watermarking Space.
In our algorithm, we propose to use the frequency transformed domain on the original medical image for integration of the watermark. First, the original image is divided on subblocks of 8 × 8 pixels. Second, applying the discrete cosine transform on each subblock of 8 × 8 pixels to obtain a subblock of 8 × 8 coefficients.
ird, the quantization step is applied to a different transformed block. Each quantized transformed block includes three parts: the first part contains the low frequency, the second encloses the midrange frequency, and the latest represents the high frequency. It is interesting to choose a medium frequency band providing good compromise between minimal degradation of the image quality and maximum resistance to several attacks. On the contrary, we apply Weber's low to the original medical image in order to obtain the suitable visibility factor of each region of the image. Our essential goal consists of enhancing the robustness of our proposed algorithm. Scientific Programming Figure 6 presents the different stages for the preparation of the space where the watermark will be hidden.

Preparation of the Watermark.
e computation steps of the preparation of watermark (information inserting) are realized as follows.
Step 1: generate a signature (sized of 160 pixels) detained by the hospital center; the principal aim is to control the authenticity of the data of patient. e SHA-1 (Secure Hash Algorithm) is useful [35].
Step 2: converting the rest of the patient's data into a binary sequence.
Step 3: the sequence and the signature are concatenated to get the message. is message will be inserted in the transformation domain of the image. Figure 7 summarizes the diverse stages to follow to prepare the watermark to be inserted.

e Embedding Process.
Our proposed approach is based on the discrete cousins transform for hiding the watermark. e embedding process, as shown in Figure 8, is described as follows.
Step 1: by using Weber's low, select the suitable visibility factor α i of each subblocks sized 8 × 8.
Step 2: the quantization step is applied on each transformed subblock sized 8 × 8.
Step 3: select the maximum value located in the midrange of the quantified transformed subblocks.
Step 4: the i th value of the binary message W is multiplied by the visibility factor α i related to the subblock to be watermarked. en, the result of multiplication is added with the maximum value of the quantified transformed subblock X i to obtain the watermarked subblock Y i .
Step 5: perform the inverse discrete cousins transform to obtain the watermarked image.
Step 6: apply the AES algorithm encryption.

Encryption Process.
e encryption step consists of a combination of the AES and chaotic techniques in goal to enhance the security of the proposed method. e desired results have been realized in three steps: Step 1: shuffling using Arnold's cat map Step 2: key generation using a chaotic Henon map Step 3: running the encryption operation using the AES algorithm e proposed encryption process supports variable length keys (128, 192, and 256 bits). It takes the employer defined secret key and the watermarked medical image as inputs and produces a scrambled, unclear, and unreadable watermarked medical image. Due to the random and chaotic key generation method, the encryption process is more complex, secure, and complicated in structure. e encryption step is easy to implement, simple, and understandable. e block diagram of the encryption process is shown in Figure 9.

e Extraction Process.
In general, the extraction step follows the reverse stages with respect to the insertion.
Step 1: using Arnold's cat map to shuffle the watermarked encrypted image.
Step 2: key generation using a chaotic Henon map and running the decryption operation using the AES algorithm Step 3: the AES algorithm to decrypt the watermarked encrypted image.
Step 4: applied the discrete cosine transform on the watermarked and original images.
Step 5: after applying the quantification step, the quantified subblocks of the watermarked and original images are carried out.
Step 6: selecting the maximum value situated in the medium frequency of each quantified subblocks. en, extract the different bits of each subblocks by using the inverse of expression (7).
Step 7: concatenate all bits extracted from different subbands to obtain the extract message.
Step 8: apply the decoder algorithm on the binary message by using the serial turbo code.
Step 9: finally, the patient's information is extracted, audited, and eventually corrected.

Performance Measures
e performance of the watermarking/encryption system is evaluated based on these following measures.

PSNR (Peak Signal-to-Noise Ratio).
A PSNR indicates that the host image and the watermarked image are more similar. In fact, this resemblance means that the message inserted is more imperceptible. In general, if the PSNR value of the watermarked image is greater or equal to 27, then the Scientific Programming PSNR is acceptable [7,36]. e PSNR is defined by the following equation: with I(i, j) and I ′ (i, j) as the pixel value of the pixel (i, j) in the host and watermarked image and A and B are the width and height of the host image.

WPSNR (Weighted Peak Signal-to-Noise Ratio)
. e WPSNR is founded on the comparison between the watermarking image and the host image (pixel to pixel) [12]. WPSNR is defined by the following equation: with var(i, j) as the local variance of pixel (i, j).

NCC (Normalized Cross-Correlation).
e robustness of the watermarking system depends on the correlation factor. Indeed, the resemblance between the original and extracted watermark is detected by the calculus of different distances among the detected and inserted watermark. is measure is performed via the NCC (normalized cross-correlation) [23,37,38]. Its value is generally between 0 and 1. Ideally, it should be 1.

e Entropy.
Entropy is a magnitude characterizing the quantity of data contained in an image. In fact, if the dispensing of grey values is very uniform, the information entropy is greater [39,40]. e entropy E(s) is defined by the following equation:   Scientific Programming

Experiments' Results of the Watermarking Scheme.
It is very essential to obtain a good value of NCC, PSNR, and WPSNR regardless of the system or the domain for dissimulated of the signature. In fact, this is particularly true in the domain of the medical imaging. For the validation of our system, we have been using diverse images sized of 512 × 512 pixels (radiographic, echographic, MRI, and mammography images). We obtain, in the literature, an acceptable extracted watermark when the value of the normalized cross-correlation is greater than "0.75" [13,41,42]. In the medical imaging domain, the signature is a binary message. So, in the phase of extraction signature, it is necessary to obtain the total of watermark inserted. Otherwise, we obtain a false diagnosis. Indeed, the measuring of the PSNR and WPSNR are acceptable when the rate of PSNR and WPSNR gives a  Scientific Programming value exceeding 35 dB [8,14,43]. We begin analyzing the experimental results by determining the PSNR, NCC, and WPSNR, as shown in Figure 11, without the application of attacks. Our algorithm will be more robust whether it will be resistant against different types of attacks. After applying several attacks, we extract the embedded signature and compare it with the original signature. e principal goal is to guarantee that the signature is not modified by the applying attacks. en, regardless of the proposed schema of watermarking, it is important to get an NCC � 1 and a good value of PSNR and WPSNR.
Among the important watermarking constraints is the robustness, next to diverse types of attacks. We have selected to apply the most attacks to analyse the robustness of the schema which are JPEG attacks, filter, Gaussian and impulsive noise, geometric transform, and cropping. en, after attacking the four types of the medical image, we attempt to extract our signature and calculate the NCC value. Our aim is to conclude the degree of robustness of our schema to resist to diverse attacks. Tables 1-6 show the experimental results of the NCC, WPSNR, and PSNR values between the host and extracted signature after applying of different attacks.

Experiments Results of the Encryption Schema.
e four medical images (IRM, radiologic, echographic, and mammographic) encrypted and transformed into an illegible 2D sequence are exposed on Figure 12. e histogram of the four encrypted and original images is also illustrated.
As the results are shown in Figure 12, the histogram of the encrypted image has a regular intensities' distribution. In fact, on comparing the histogram of the watermarked encrypted image with the histogram of the host image, we find that the probabilities of occurrence of every grey level are uniformly repartees. e computed correlation coefficients were very low, and there was low similarity among the pixel values when the enhanced AES encryption was utilized under the CBC mode of operation. erefore, the experimental results recommend that the encryption process can generate reliable results and is highly secure for medical imagery with a uniform distribution and maximum dissimilarity of pixel values without leaving any clue for statistical analysis and attacks. erefore, the entropy of the encrypted watermarked image is high (close to 8 bits/pixel). e entropy of the encrypted and watermarked medical images is shown in Table 7.
We evaluate the algorithm for four images. For each image, we produce the correspondent encrypted watermarked image utilizing the encryption-system. Subsequently, we evaluate it by calculating the entropy value. Compared with recent works in [44][45][46][47], as shown in Table 8, we obtained a good entropy very close to 8 bits/pixel.

Discussion of the Proposed Schema.
In this section, the proposed system is studied by comparing our results to those cited in the section of a review of related works. Comparing the psycho-visual quality of the original and the watermarked medical image, the proposed algorithm provides very good results. In the absence of any type of attack, the PSNR equal to 59 Table 9 illustrates the NC and PSNR values of our proposed system computed according to four medical images and compared with five recent schemes in attack-free case [48][49][50][51]. It can be observed that the NC values are equal to 1 and the average of the obtained PSNR values (61,7769 dB) is moderate. e obtained results are much higher than the reported methods. erefore, our proposed scheme performs more efficiently depending on the imperceptibility factor.
During the manipulation of watermarked decrypted medical images, these later might be altered by various voluntary and involuntary image processing attacks. Consequently, it is essential to evaluate the robustness of our proposed approach under an attacking environment. For this purpose, a set of four medical images watermarked are subjected against the following attacks: JPEG compression, Gaussian noise, median filter, cropping, and rotation. After applying those attacks, it is necessary to evaluate the PSNR,        Table 1 show that, from a compression rate equal to 60%, we were able to correctly extract our watermark (NC � 1). Out of this value, the recovery of the watermark is not complete, and then, there will be an erroneous diagnosis. Detection by correlation between the extracted and inserted watermark made our approach robust against several attacks, while keeping the visual appearance of the image after watermarking. To evaluate our method against the Gaussian noises, we obtained very promising results. As the results already presented in Table 2, the approach that we have proposed has proven its performance. Indeed, the recovery of the watermark is equal to 100% (the signature is extracted correctly from the 500 signatures) for a variance equal to 0.07; beyond this value, the recovery of the watermark is not total. We validate the results obtained above that our approach has proven its robustness against this type of attack.
We also tested our system against salt and pepper attack. It is remarkable, as the results presented in Table 3, that data recovery is complete until a density value equal to 0.07. Beyond this value, the recovery of the watermark is not complete. As the values of PSNR and WPSNR, the results have proved that we have kept the image quality during watermarking. e detection by correlation between the signature extracted and that inserted argued the robustness of our approach against salt and pepper attack.
We are testing our system against the median filter [52,53]. e test is evaluated with various sized windows (from (3 × 3) until (9 × 9)) ( Table 4). In fact, the recovery of

18
Scientific Programming our signature is equal to 100% (extracted correctly the watermark in the presence of this attack type) up to coefficients (6 × 6). e detection by correlation between the extracted and inserted signatures made our approach robust against attack by a median filter, while keeping the visual appearance of the image after watermarking.
We tested the robustness against the rotation attack. For a value of rotation angle greater than 10, the recovery of the    watermark is not complete, and then, there will be an erroneous diagnosis. We also tested our method against the cropping attack. As the results presented in Table 6, the approach that we have proposed has proven its performance. In fact, the recovery of the watermark is equal to 100% (NC � l) up to a window size (64 × 64); beyond this value, the recovery of the watermark is not total. We validate the results obtained above that our approach has proven its robustness against this type of attack. Table 10 gives the results compared with those obtained with the approaches described in papers [49][50][51].

Conclusion
e present work is a new robust joint encryption/watermarking algorithm using the discrete cosine transform and the advanced encryption standard. e main contribution of this manuscript consists in improving the two principal factors existing in all watermarking systems (imperceptibility and robustness). In order to enhance the factor of imperceptibility, we chose to hide the watermark in the medium frequency of the DCT Transform. Weber's law is used to determine the suitable visibility factor for each subblocks of the image. To do so, we came across several stages. e original medical image is subdivided into subblocks sized 8 × 8. After that, the DCT is applied on each subblocks, and the quantization phase is then used in order to determinate the space of watermark insertion. e AES encryption algorithm is used together with the serial turbo code to improve the robustness of our contribution. We obtained a good result in terms of the watermark extracted which is comparable to the original (NC � 1). To evaluate the performance of our system, the proposed method is applied on medical images. Various tests are performed, such as JPEG compression, geometric transformation, and adding noise. e results show that our proposed algorithm is very robust against these various types of attack [40,[54][55][56][57][58][59].

Data Availability
e obtained results used to support the findings of this study are included within the article.

Conflicts of Interest
e authors declare that they have no conflicts of interest.