Digital watermarking is the most commonly used technique for securing the data against attacks. Visible or Invisible watermarking are the two different types of watermarking approaches used ^{1}. The watermarking performed defines how thewatermark is inserted. It is important to use an appropriate decoder for detecting the existence of a watermark in an image. There are two different ways to explain the watermark with the object and human perception. Depending upon the object means text, image, audio, and human perception means visible and invisible watermarking. The sensitivity of the visible watermark is high and any kinds of changes seen within the marked image are detected through it. SemiFragileWatermark is invisible but is fragile to malicious modifications. However, it has robustness towards the incidental manipulation due to which image authentication process is including it. RobustWatermark can hear the communication attacks even though it is invisible. Two main factors in a digital watermarking system are a watermark embedder and a watermark detector. The watermark is inserted on the cover signal using a watermark embedder. Further, the existence of a watermark signal can be detected by the watermark detector. The process through which image watermarking is performed at the source end is known as the watermarking embedding process. A watermarking algorithm is a procedure to embed the watermark within the host image ^{2}. The embedding algorithm is reversed for extracting the watermark from the watermarked image.
In the Proposed work a review of existing digital watermarking methods with various techniques. The methods of digital image copyright are becoming popular day by day. It has embellished all the demands in the field of multimedia objects to restrict the approach of unauthorized manipulation and duplication of original digital objects ^{3}. The digital revolt in digital image processing has made it possible to create, manipulate, and transmit digital images in a simple and fast way. The opposing effect of this is that the same image processing techniques can be used by hackers to interfere with any image and use it criminally ^{2}. By dividing the real image into a cover image with the help of the Wavelet Transform and Finescale DWT coefficients are represented by (LL, LH, HL, HH) while coarsescale DWT coefficients are constituted by frequencybased LL
A technique for embedding the watermark in the host image such that the best image quality can be achieved ^{22}. There is no direct insertion of the watermark bits into the frequency coefficient when the watermark bits are inserted into particular frequencies of the image. For providing additional security, it is important to scramble the watermark before embedding ^{23}. BER, a performance metric is used for the decoder evaluation and comparisons are made with existing decoders and their proposed decoder ^{24}. YCbCr color space for watermark embedding with A human visual system (HVS) nonblind watermarking scheme. The new algorithm has been referred to as the Additive Embedding Scheme (AES), The embedding factor for each module is considered with the less noticeable falsification and the singular vectors of the HL subband of DWT. The PSNR and NCC performance metrics of the extracted watermark are evaluated with robustness and transparency ^{25 26.}
The Color image is used as watermarking as an alternative to the binary or gray image. The YCbCr color space is used to amplify the correlation between the original and the watermark image. Arnold transform has been used for secure watermark. The authors workout to an imperceptibility and fidelity with NCC, PSNR, and SSIM parameters. The number of attacks has been done for testing and robustness ^{27}. The random diffusion process overcomes the incomplete period of mapping. Iterations of Arnold transform are dependent on a secret message that means with low significant bits of the cover image can effectively recover the original image. A novel image encryption algorithm based on bitlevel Arnold transform and the hyperchaotic map has been proposed ^{28}. To resolve the old encryption algorithm with a new concept that is called VMIE. VMIE is the visually meaningful image encryption method used to avoid human eye detection with a secure encryption algorithm on color images. The technique of DWTDCT and SVD was performed with YCbCr color space. Qihyper chaotic method is used for preencryption. The VMIE scheme mainly works on common attacks^{ 29.}
The algorithm known^{ }as SAES introduced to overcome the directional features and imperceptibility and also the robustness of the image. Subsampled shearlet transform is used to improve the watermark algorithm with antigeometric attacks authors proposed research mainly works on the imperceptibility and robustness by embedding procedure used to resolve the false positive problem ^{30}.
CA and DCT are used for embedding the segments of the image to define the behavior of the image. The color component Y from YCbCr color space has been used and the super pixel of an image is defined as homogenous and heterogeneous blocks. This method is applied with DCT and CA by embedding process in Cb, Cr color components. The Arnold transform is used for security purposes. Authors works on state and art experiment method for comparison and better results^{ 31, 32. }The authors worked on YCbCr color space with double encryption. They used the sensor nodes as compares to original image. PSNR of this work is greater than 40 dB and the NCC is also high. The authors proposed work had been on DCTDNA and chaotic encryption methods. The provide copyright protection, authentication and tamper localization of color images.
In this paper, we have presented a system that uses the dual encryption scheme and DWT and SVD. We have also proposed a dual encryption scheme through Arnold’s transformation and chaos map in YCbCr color space. Starting with the system we can define the first half of the system with a dual encryption scheme & channel coding and conversion of RGB to YCbCr color space after that second half defines 4 levels of DWT with GLCM and SVD. The cover watermarked image is encoded and the singular value decomposed by SVD. The four levels of DWT are applied to it, and the singular value matrixes are embedded into the Y, Cb, Cr components of the host image. The embedding factor for each component is calculated with singular vectors of the hl subband of DWT with bit selection automatically by GLCM. In this paper, the GLCM technique is used for better results for enhancing the performance of the watermarked image affected by degradation with the DWT method. The various existing methods working for digital watermarking are being done in the introduction part. The continuing section of this paper is organized in the following section. Section 2 presents the Materials and Methods. Section 3 discusses the Performance evaluation. Section 4 presents the results and discussion of the proposed work. Section 5 concluded this paper.
From the introduction section discussed above, the following problems of digital watermarking are identified. These problems are considered as challenges and addressed by the proposed method.
YCbCr color space is not used by several authors, we have used this color space for the proposed work.
Several authors used one encryption either Arnold transform or chaotic map, we proposed the dual encryption scheme.
DWT technique is used mostly but With DWT embedding the degradation in watermark image in the frequency domain. We have to use the GLCM method to increase feature extraction.
Channel coding used in watermarking is a complex task but we have applied FFT that has Eased to use.
Different watermarking techniques are applied and performance measures either by numeric values or structural features. We have work on both, numeric and structure feature measurements with PSNR and MSSIM, BER is used to measure the channel coding errors in the proposed work.
From
We propose algorithms embedding and extraction based on dual encryption and channel coding methods from
A Robust Watermarking Scheme in YCbCr Color Space Based on Channel Coding
^{32 }DCTDNA and chaotic applied on but the PSNR value low, we apply a chaotic map and Arnold transforms in image encryption and security basis with DWTSVD.
References Year Description Outcomes 2014 A lossless digital watermarking approach is proposed in which zeroperturbation is applied to the digital image maps which include content and graphics in them. The authors proposed an approach operated within the redundancy regions of maps and provided higher scalability to the topology changes as per the results achieved through the comparative analysis of their proposed and existing techniques. 2014 FIFT and inverse FIFT used for channel coding. In which the author defines OFDMMIMO techniques for channel coding 2015 A novel digital watermarking algorithm is proposed within NSCT domains to be applied in the copyright protection field. By authors,' good invisibility, robustness, and capacity are achieved by their proposed approach as per the simulation results. The commonly found image processing and combo attacks are also resisted effectively by their proposed approach. 2015 An invisible grayscale logo watermarking is proposed in which the logo is enhanced using adaptive texturization. The watermarking task is to be recast into a texture similarity task by applying this proposed approach The performance of the author's proposed algorithm is to be better as compared to existing approaches as per the tests performed with multiple logos on a dataset of host images and in the presence of several types of attacks. 2015 A novel watermarking algorithm is proposed based on the partial pivoting of lower and upper (PPLU) triangular decomposition. The reliability of the author's proposed algorithm is shown higher along with improvement in imperceptibility against the existing techniques as per the conducted experiments and achieved results. 2016 A prediction error expansion based watermarking mechanism is thus proposed here. Promising results are achieved as per the comparisons made amongst the authors' proposed and existing techniques. 2016 A blockbased mechanism is proposed in this paper using SVD and DCT along with the human visual system. To choose significant blocks for embedding the watermark, entropy and edge entropy are utilized as HVS properties by the proposed mechanism. As compared to existing approaches, the performance of the author's proposed approach was shown to be better. The AES192 was applied for encrypting an area of important information such that the security problem was improved. 2017 An improvement is proposed for the existing watermarking approaches by proposing a novel approach. The watermarks are simulated successfully and are applied to five various watermarking approaches as per the experiments performed by the authors. 2018 Authors proposed a novel approach that could be applied to color as well as grayscale images that was based on chaotic encryption and was named as a blind digital image watermarking approach. For most of the image processing operations, the authors' proposed approach has provided higher robustness as per the simulation results. It is seen that with respect to security, imperceptibility, and robustness, the performance of their proposed mechanism is better. 2018 The authors proposed a novel approach for medical applications to provide image authentication and selfrecovery. The image tampering is localized and the original image is recovered by proposing this new fragile watermarkingbased approach. The author’s experimental results achieved that the tamper location accuracy and PSNR of the selfrecovered images are enhanced to a higher level as compared to the existing approaches. 2018 A new embedding domain is proposed for blind image watermarking which was known as Discrete Shearlet Transform (DST). Authors Comparisons are made against Discrete wavelets and Contourlets which show their proposed technique provides higher windowing flexibility with higher sensitivity to the directional and anisotropic features. 2018 A novel approach is proposed using optimal DCT psychovisual threshold such that high imperceptibility and robustness can be provided to protect the copyright of the image. The performance of the author's proposed technique is better in terms of robustness and invisibility as compares to existing. During the presence of various types of attacks, high image quality is achieved through watermark extraction. 2018 Hybrid scheme is used in YCbCr color space with Arnold transform Security, fidelity, imperceptivity and robustness by NCC, PSNR, SSIM parameters 2020 A new concept VMIE is used to avoid human eye detection with secure encryption by Qihyper chaotic Security and test common attacks 2020 Sub sampled shearlet transform, SAES and false positive problem resolved Directional features, imperceptibility, robustness with various attacks 2020 Cellular automata and DCT on super pixel are the new method used for embedding segments and Arnold transform Security and state & art experiment on number of attacks for better results
This section explains the proposed hybrid watermarking scheme using dual encryption and channel coding with the YCbCr color space. The dual encryption and channel coding algorithms are designed by a chaotic map, Arnold transforms, and fast Fourier transform. From ^{26, 28} and various encryption references given in related work, we design and experiment with the results by various performance metrics like PSNR, NCC, etc. Different types of attacks are applied to test images and the best results are found in the proposed work by MATLAB 2016b simulator.
YCbCr is a kind of linear color space, in which Y denotes the luminance module and Cb and Cr are the attentiveness modules of blue and red ^{26}. RGB image can be converted to YCbCr color space by following Equation 1(ac):
Reversibly, YCbCr color space to RGB image conversion is as follows in Equation 2(ac):
The color sensitivity of the human visual system, the Cb channel is the smallest sensitive. It means that the scheme of embedding watermark information in the Y channel is more robust, while watermark insertion in the Cb channel has good transparency.
The behaviors of a particular nonlinear dynamic system that represents dynamics that are sensitive to initial situations are defined by chaos theory. Sensitivity to initial conditions and mixing property are the two important properties of this approach. Enormous deviations are caused by the corresponding orbits due to the small deviations in the initial conditions. Thus, for the inflexible chaotic systems, a longterm forecast is rendered. For entropy production, this deterministic is a local mechanism that exists in principle but is not determinable in dynamic behavior. Also, Entropy producing deterministic systems is the other name for chaotic systems. From the available memory, the number of steps also known as the horizon of predictability is increased. The architecture of the chaosbased image encryption method is shown in [
An image scrambling method used for the image data can be encrypted and decrypted is known as Arnold transform. Missing any information, this transform is the areapreserving and invertible. To unclear the image outside recognition, it is possible to perform mapping a number of times. The mapping can be done successively many times to completely unclear the image beyond recognition
Here,
To represent essential properties of an image, linear geometric technique known as SVD is used from ^{27}. SVD is Singular value decomposes; It Decomposes an image represented by m × m matrix (C) into two orthogonal matrices (U_{c} and V_{c}) and one diagonal matrix (S_{c}) whose entries are known as singular values of the matrix C. This type of decomposition is called singular value decomposition of C and can be expressed in equation 4:
Where, U_{c} is a m × m matrix with orthogonal columns consisting of lefthanded singular vector, S_{C} is an m × m diagonal matrix having nonnegative singular values as a diagonal element arranged in descending order and V_{c} is an m × m matrix with orthogonal columns known as righthand singular vectors. Use of SVD in digital image processing field has number of benefits like;
Singular values are more robust against various operations of image processing and attack.
The size of the matrices is variable like a square or rectangle.
Larger singular values not only preserve most energy of an image but also show the resistance against various attacks.
As many small singular values of the S matrix reflect geometrical features of the image, hence, minor variations in singular values of an image does not produce any noticeable change in the original image.
DWT is an ordered transform. The multiresolution analysis is given with DWT signals. In DWT the signals are divided and pass onto the high and lowfrequency subbands. Highfrequency subband contains information regarding edges and human eyes less sensitive to the changes on edges. There are different types of wavelet functions like Daubechies, Morley, Marr, Harr, etc. This transform mainly highlights small waves called wavelets with fluctuating frequency and inadequate duration. Wavelet transform gives spatial and frequency description of an image. DWT works on 2D images and is processed by 2D filters on each dimension. Wavelet filters divide into four nonoverlapping multiresolution subbands like LL, HL, LH, HH. The LL subband defines the coarsescale coefficients and the other subband defines the finescale coefficients. Embedding in lowfrequency subbands could increase the robustness, and highfrequency subbands include texture and edges of the image with the human eye is not sensitive to modification in these subbands^{ }
Subbands consist of a wide range of the frequency spectrum of the image. Therefore, the robustness of the watermarking system is increased. A 4level DWT is shown in [
Channel coding is used to protect digital information from noise and interference and reduce the number of bit errors in the digital communication system. Channel coding is typically capable of selective introduced redundant bits into the transmitted information stream. These additional bits drive for allow detection and correction of bit errors in the received data stream and provide more consistent information transmission. In the proposed work we can use fast Fourier transform as a channel coding.
A faster version of the Discrete Fourier Transform algorithm in which few highly efficient algorithms are used which perform similar actions but at very high speed is called FFT. Since a discrete signal from the time domain is transformed into discrete frequency domain representation, DFT is considered to be highly important during frequency analysis.
Here, the numbers of harmonics are represented by n, the period of the signal is represented by N, and the nth harmonic is represented by n/N.
A way in which the secondorder statistical texture features can be extracted is known as GLCM. GLCM is a wellestablished statistical device for extracting secondorder texture information from images. This technique introduced by haralick in which two steps for feature extraction are proposed. In the first step of computing, the cooccurrence matrix and instep are calculating texture features based on the cooccurrence matrix. This technique is mostly used from biomedical to remote sensing image analysis. A GLCM is a matrix where the number of rows and columns is equal to the number of distinct gray levels or pixel values in the image of that surface. GLCM is a matrix that describes the frequency of one gray level appearing in a specified spatial linear relationship with another gray level within the area of investigation.
This research work is based on the generation of watermark images using the blockbased method. In the proposed method, an image is taken as input which can be divided into a certain number of blocks. The image which is divided into a certain number of blocks can be divided into 8*8 blocks each. The technique of DWT is applied which can calculate wavelet features of each block. To perform image embedding, a discrete wavelet transformation mechanism is applied. An inefficient approach is chosen randomly for image embedding which is bit selection. There is degradation in the quality of the watermark image when randomly selecting the bits. To dynamically choose the embedding bit, this research applies the Gray Level Cooccurrence Matrix algorithm. The watermark is taken as input which can be processed with the chaos encryption scheme. The chaos encryption scheme needs the transformation which is transformed using the Arnold transformation method. The technique of channel codes scheme is applied which can generate the codes. The technique of embedding will be applied which can generate the watermarked image. The extraction method is performed with decryption and inverse algorithms. The proposed technique is the hybrid scheme as it uses the encryption scheme and also transformation techniques to generate a watermarked image with channel coding in YCbCr color space. In the Proposed watermarking scheme have two procedures, the first is the embedding procedure and the second is the extraction procedure.
In the Embedding procedure, an image is taken as input which can be divided into a certain number of blocks with preprocessing work. The image which is divided into a certain number of blocks can be divided into 8*8 blocks each. The technique of DWT is applied which can calculate wavelet features of each block, after applying GLCM the degradation with DWT is eliminated and the inverse IDWT is applied for the watermark image. For the security of the watermark, we can apply dual encryption and FFT
There are some steps given below to define the proposed embedding procedure:
Read the host image with the size of [500,500].
Convert RGB image into YCbCr color space
Read watermark image and perform chaos encryption n time on each block.
Apply Arnold transform on each block.
Perform channel coding
Perform SVD and 4 levels of DWT with GLCM on the YCbCr components of the host image. Sub band components LL, HL, LH, and HH are accomplished.
Apply Singular value matrix to hl of the fourth DWT.
Calculate watermark strength factor for different host channels Y, Cb, Cr
Embed watermark information with host components Y, Cb, Cr.
Update all the hl subband components
Apply inverse DWT for host components
Convert watermark image to RGB image.
[


Box 1: 

For i=1: m do 

For j = 1: n do 

Sort the chaotic sequences and using it for permutator 

of row by x sequence; 

if the chaotic x sequence is odd then 

Circular shift row pixel to the left 

End 

else if the chaotic x sequence is even then 

Circular shift row pixel to the right 

End 

End 

End 

for i=1: n do 

for j = 1: m do 

Sort the chaotic sequences and using it for permutator 

of row by y sequence; 

if the chaotic y sequence is odd then 

Circular shift column pixels downwards; 

End 

else if the chaotic x sequence is even then 

Circular shift column pixels to the upwards 

End




host image (I) 

Public key alpha (α) 

set of textured blocks (B) 

Output: Watermarked Image (Iw) 

Start: 

1 Partition I into L X L blocks 

2 for each blocki ϵ I do 

if block ϵ B then 

blockiw= blocki+ α 

Else 

blockiw= blocki 

end if 

End loop End 

Fourier Transform is defined as decomposing an image into its real and imaginary components. These components are used for a representation of the image in the frequency domain. The input signal is an image at that point the number of frequencies in the frequency domain is equivalent to the number of pixels in the image or spatial domain. The inverse transformation is the retransforms the frequencies to the image in the spatial domain. The FFT and its inverse of a 2D image are shown in equation (67):
Where f(m,n) is the pixel with coordinates (m, n), and F(x,y) is the value of the image in the frequency domain equivalent to the coordinates x and y, the M, and N are the dimensions of the image. The implementation requires the dimensions of the image are in the power of two and the transformation of N points can be written through the sum of two N/2 transforms (divide and conquer method). The result of the Fourier Transform is a complex number and has a greater range than the image in the spatial domain.
The polynomial is operation which is performed on the data. “ω” has the value Zero. The fast Fourier transform is a method that allows computing the DFT in O (nlogn) O (nlogn) time. The basic idea of the FFT is to apply divide and conquer. We divide the coefficient vector of the polynomial into two vectors, recursively compute the DFT for each of them, and combine the results to compute the DFT of the complete polynomial
Input: Coefficient representation of a polynomial A(x) of degree ≤ n − 1, where n is a power of 2
Output: Value representation A(ω^0), . . . , A(ω^n−1)
if ω = 1: return A (1)
express A(x) in the form Ae(x^2) + xAo(x^2)
call FFT (Ae, ω^2) to evaluate Ae at even powers of ω
call FFT (Ao, ω^2) to evaluate Ao at odd powers of ω
for j = 0 to n − 1:
compute A(ω^j) = Ae(ω^2j) + ω^jAo(ω^2j)
return A(ω^0), . . . , A(ω^n−1)
Given an image, each with intensity, the GLCM is a tabulation of how often different combinations of gray levels cooccur in an image or image section. In GLCM the cooccurrence matrix is computed based on two parameters, which are the relative distance between the pixel pair d measured in pixel number and their relative orientation theta. Normally theta is quantized in four directions i.e., 0, 45, 90, 135, even though various other combinations could be possible. GLCM has fourteen features but between them, the most useful features are ASM (), contrast, CORRELATION, inverse difference moment, sum entropy, and information measures of correlation. Each element (i,j) in GLCM specifies the number of times that the pixel with the value i occurred horizontally adjacent to a pixel with value j. [
The GLCM texture considers the relation between two neighbouring pixels in one offset, as the secondorder texture. The gray values relationships in a target are transformed into the cooccurrence matrix space by a given kernel mask such as 3*3, 5*5, and so forth. In the transformation from the image space into the cooccurrence matrix space, the neighbouring pixels in one or more or some of the eight defined directions can be used; normally, for directions such as 0, 45, 90, and 135 is initially regarded, and its reverse direction (negative direction) can be also counted into account. It contains information about the positions of the pixels having similar gray level values.
The watermark extraction process is reverse to the embedding process. Firstly, the watermarked image is converted to YCbCr color space, and 4level DWT is performed. The singular values matrix of the watermark is extracted from the singular values matrix of the channel components. We calculate the encoded bits of the watermark and reshape them to bits sequence as the input of the channel decoder. The decoded bits are reshaped and Arnold transformed and chaotic maps applied inversely, then the extracted watermark is achieved. In the proposed paper inverse of Arnold transform and decryption of chaotic map and channel, decoding is explained by algorithms. [
The complete stepbystep extraction process is discussed below:
Read the host image with the size of [500,500].
Convert RGB image into YCbCr color space.
Read watermark image and perform chaos encryption n time on each block.
Apply 4 levels of DWT on Y, Cb, Cr components of the watermarked image, and subband components LL, HL, LH, and HH are accomplished.
Apply Singular value decomposition to hl subband of the fourth DWT.
Extract watermark information with host components of y, Cb, Cr.
Calculate encoded bits with chaotic map
Apply channel decoding
Embed watermark information with host components Y, Cb, Cr.
Reshape them according to decoded bits
Apply the inverse of the Arnold transform.
Perform decryption of a chaotic map and achieve the extracted watermark image.
The extraction procedure follows the Decryption of a chaotic map, Inverse of Arnold, and Inverse of Fast Fourier transform algorithms explained (4,5,6).
Image is the variable which store transformed image
Input: Transformed Image
Output: Detransformed Image
for inc=1:num
For irow=1:irown
For icol=1:icoln
inrowp = irow;
incolp=icol;
For nite=1:inc
inewcord = [2  1; 1 1]*[inrowp, incolp];
inrowp=inewcord (1);
incolp=inewcord (2);
End
iminverse (irow, icol) =newim ((mod(inrowp,irown)+1),(mod(incolp,icoln)+1));
End
End
Input: Encryption Image C and Secret Chaotic Keys
(a, b, X0, Y0), where a and b are constants.
RI is the variable which store value of reconstructed image. CDC value store lowest pixel value. The CAC is the variable which store swiped blocks. The Henon is the variable which store map sequence. The key is generated after applying the chaotic key generation steps.
Output: The Reconstructed Image (RI)
Separate pixel of C into the lowest pixel in CDC and CAC
according to the inverse chaotic swapping: (CDC, CAC) = Chaotic Swap (C).
Generate Chaotic Sequence according to the Henon map:
Convert the sequence Xi and Yi into an integer value.
Decrypt CDC and CAC using Chaotic Decryption:
AC = Chaotic Decryption (CAC, Y) Decrypt CDC using RC4 by Secret KeyX:
DC = RC4_Decryption (CDC, X)
Compute the inverse of DWT
Output RI.
Input: The Encoded image
Output: The output is decoded image
(c, s) = (w.real, w.imag)
a = np.array([x.real, x.imag]) /// The image is the twodimension array which Store image
if s == 0:
Pass
elif c >= 0.0:
a[0] = int(a[1]*(c1)/s)
a[1] = int(a[0]*s)
a[0] = int(a[1]*(c1)/s)
else:
a *= 1
a[0] = int(a[1]*(c+1)/s)
a[1] = int(a[0]*(s))
a[0] = int(a[1q]*(c+1)/s)
return complex(a[0], a[1]) // The complex is the function name which return decoded image
From the challenges of an image watermark, the two most factors affect the image watermark. one is the quality of the watermark and another is the security of the watermark. Performance evaluation is the main concern to exploits the output from various techniques. We can use different types of performance metrics like PSNR, NCC, MSE, etc and different types of attacks are applied to color images to perform various metrics.
MSE is an image watermarking used to calculate the middling of the squares of the errors between the host and watermark images. It is defined with E. The main limitation of this metric is that it depends strictly on the numeric comparison. It means no level of the biological factor of the HVS measures. for an M*N twodimensional image, the computation formula shown in equation 8:
Where E(i, j) denotes the pixel value of the host image, and E'(i, j) denotes the pixel value of the watermark image. The value of MSE is high means the larger the distortion caused by the watermark and the attacks.
PSNR is the estimation between the host image and the watermark image. In general, the peak signal to noise ratio is the ratio between the maximum power of the signal and the power of distorting noise that affects the quality of signal representation in image extraction. The PSNR is generally defined with a logarithmic decibel scale. The higher the PSNR value is lesser the difference between the host image and the watermark image, which means improved watermarking transparency. The dimensions of the correct image and degraded image matrix should be indistinguishable. For an M*N twodimensional image, the computation formula of PSNR is shown in equation 9:
PSNR value is calculated based on MSE.
To measure the robustness of the watermarking technique normalized correlation coefficient method is used.NC is the correlation between the original watermark and the extracted watermark in the digital watermarking procedure. The NC formula is described in equation 10 to calculate the watermark coefficients for M*N twodimensional image. W is the original image and w' is an extracted watermark.
BER is expressed as a ratio, it means the number of erroneous bits received over the total number of transmitted bits. The bigger the BER value, the poorer the performance of the system. if the medium between the transmitter and receiver is good and the signaltonoise ratio is high then BER is very small that has no perceptible effect on the overall system; if noise can be detected then there is a chance for BER will need to be considered. BER is mainly used to measure channel errors. The computation formula is shown in equation 11:
It usually works on the HVS (Human visual system) based measurement. It is based on SSIM. The overall image quality MSSIM is obtained by calculating the mean of SSIM values over all windows as in equation 12:
Where p is the number of sliding windows.
The steps are followed for the computation of MSSIM are:
The host and distorted images are divided into blocks of size 8*8 and then the blocks are converted into vectors.
Two means and two standard deviations and one covariance values are computed from the images as in equations 1317:
Where µ_{x}µ_{y} denotes the mean values of host and distorted images and µ_{x}µ_{y}_{ }denotes the standard deviation of host and distorted images, and µ_{xy} is the covariance of both images.
Luminance, Contrast, Structure is described in equation 1820 and a comparison based on statistical values is also computed.
The structural similarity index measures between image x and y are in equation 21:
Where c1, c2 are constants. SSIM applied locally using a sliding window of size B*B that moves pixel by pixel horizontally and vertically covering all the rows and columns of the image, starting from the left top corner of the image. MSE strictly computes numeric values, no structural features measures, but MSSIM gives a solution for structural feature detection performance metrics.it is based on HVS that have structure quality measure of the images.
This section discusses the details about the test images used for the proposed work, parameterwise results achieved, and their comparative analysis. In this section, we evaluated the numeric as well as the structural features results with performance metrics.
There are several experimental results are given in previous research work. We can describe these results with performance parameters that are explained with performance metrics. The proposed paper explains the size of the dataset, type of dataset with color test images are tested. We can test Baboon, Cable car, Pens, Barbara, Boat to measure the proposed work performance. We can use Contrast, Salt & pepper, and Gaussian attacks. [
For experimentation of the proposed work, we have developed a simulation environment and analysed the different parameters used in digital watermarking for robustness and security purposes. We have used the girl image (shown in [
The extracted watermark (shown in [
Image Name 
Pixel Size 
AES Algorithm 
Proposed Algorithm 
Baboon 
500*500 
38.71 
42.89 
Cable Car 
500*500 
52.07 
55.78 
Pens 
500*500 
54.35 
57.89 
Barbara 
500*500 
53.56 
56.89 
Boats 
500*500 
55.94 
58.78 
Flower 
500*500 
55.79 
59.89 
Attack 
AES Algorithm 
Proposed Algorithm 
Contrast Attack 
0.934 
0.96 
Salt & Pepper 
0.956 
0.97 
Gaussian Attack 
0.945 
0.96 
As shown in [
BER has been calculated by using equation 11. The lower the BER value, the higher the performance of the system. BER is mainly used to measure channel errors. The original image used was a girl image (refer [
We have compared the proposed method with the AES method based on various parameters. For some parameters, results are not available so we have not considered those parameters in a comparative study. [
Parameters 
AES Method 
Proposed Method 
MSE 
NA 
30.78 
PSNR 
38.71 
42.89 
NCC 
0.93 
0.96 
BER 
NA 
0.56 
MSSIM 
NA 
0.48 
The digital watermarking technique is used to increase the robustness, imperceptibility, and security of digital data. We have studied the various existing techniques and applied four levels of DWT with dual Encryption. Due to the multiresolution characteristics of DWT, the proposed scheme also provides robust watermarking for images. Robustness, Imperceptibility, and Security are the three main requirements of today's watermarking system which are hard to achieve simultaneously. In the proposed hybrid scheme, we have applied the dual encryption with chaos map and Arnold transform that provides security over digital images. PSNR and NCC performance metrics are used to compare the numeric metrics but MSSIM is used to define mean structure similarity between the original and watermark image. PSNR provides absolute error onto RGB and chromatic values of YCbCr. In the proposed work, performance metrics, Attacks, Encryption, Digital watermarking techniques through embedding and extraction, etc concepts have been defined.
In previous research work, different techniques were used for security, robustness, and imperceptibility for YCbCr color space. We have reviewed various methods and we found that DWTSVD is mostly used for embedding procedure but the drawback is degradation in image quality. To overcome the degradation problem proposed research includes the GLCM algorithm for better results. Arnold or Chaotic maps was used by the number of researchers for security purpose. But proposed research work is on dual encryption for security purposes.
[
In the proposed research we have used the channel coding method through fast Fourier transform. Proposed research gives results on Numeric and also structured quality of an image. Different parameters used and compared with existing work explained in [
Sr.no 
Parameters 
SWT 
VMIE 
AES 
SAES 
CA 
Proposed 
1 
MSE 
NA 
NA 
NA 
NA 
NA 
30.78 
2 
PSNR (dB) 
37.93 
42.428 
38.71 
38 
46.53 
51.39 
3 
NCC 
0.99 
0.99 
0.93 
0.99 
0.99 
0.96 
4 
BER 
NA 
0.43 
NA 
0.20 
NA 
0.56 
5 
MSSIM 
SSIM=0.99 
NA 
NA 
NA 
NA 
0.48 
Sr. no 
Reference 
Embedding (Blocks/AES) 
Technique Used 
Encryption method 
Channel Coding 
1 

Blocks 
DWTSVD 
NA 
NA 
2 

Blocks 
DWT, embed binary 0,1 
NA 
NA 
3 

Blocks 
DCT, embed binary 0,1 
Chaotic map 
NA 
4 

Blocks 
DWTDCT 
Arnold Transform 
NA 
5 

AES 
DWTSVD 
NA 
NA 
6 

Blocks 
SWTSVD 
Arnold Transform 
NA 
7 

CCAES 
DWTSVD 
Arnold Transform 
Convolutional Code 
8 

Blocks 
DWT, DCTSVD 
Qihyper Chaotic, VMIE 
NA 
9 

SAES 
DWTSVD, NSST 
Arnold Transform 
NA 
10 

Blocks 
DCTCA 
Arnold Transform 
NA 
11 

Blocks 
DCTDNA 
Chaotic map 
NA 
12 
Proposed 
Blocks 
DWTSVD, GLCM 
Arnold Transform, Chaotic Map 
FFT 
This research is based on the image watermarking technique in which the data is secured and the data is sensitive images. To generate a watermark image, the DWT approach chooses the bit manually. To create a watermark image, the embedding bit is chosen automatically by GLCM in this research. For security purposes, the dual encryption method is used for best results with a fast Fourier transform. With different performance parameters, a comparative analysis is performed. The achieved outcomes show that when applying dual encryption with the GLCM algorithm for bit selection, around 10 to 15 percent of improvement in the results. The detailed analysis of existing work and proposed work is concluded in [
YOP/Reference 
Parameters 
Parameters Values 
Quality Metrics 
Technology Used 
2012 
1. PSNR 
DWT=58.39 dB 
Recovery of watermark 
DWT, DWTDCT 
DWTDCT=51.46 dB 

2015 [12] 
1.PSNR 
29 to 44 dB 
Imperceptibility, Robustness 
Arnold, ALTMARK 
2.NCC 
0.52 to 0.97 

2016 
1. PSNR 
8.03 to 9.68 dB 
Data Hiding 
2D Arnold Cat Mapping random diffusion 
2.MSE 
94.29 

2017 
1.PSNR 
40 to 57 dB 
Content authentication, Privacy protection, Imperceptibility 
DWTDCT 
2.NCC 
0.99 

3. BER 
0.49 to 7.97 

4.SSIM 
0.99 

2018 
1.PSNR 
51.71 to 52.19 dB 
Robustness, Perceptual Quality 
4 level DWT, Arnold, HVSJVD 
2.State of Art comparison 
AES0.99, MES0.93 

3.NCC 
AES=0.95, MES=0.97 

2018 
1.PSNR 
51 to 65 dB 
Robustness, Security, Imperceptibility 
SWTSVD, Arnold Transform 
2.NCC 
0.52 to 0.59 

2020 
1.PSNR 
42.94 dB 
Security, Imperceptibility 
DWTDCTSVD, chaotic map 
2.NCC 
0.84 to 0.98 

3.MSSIM 
0.98 

4.BER 
0.44 

2020 
1.PSNR 
38 dB 
Robustness, Imperceptibility 
DWT, SVD, Arnold 
2.NCC 
0.99 

3.BER 
0.020 

2020 
1.PSNR 
44.5 dB 
Security 
DCT, CA, Arnold, 
2.NCC 
0.98 

2020 
1.PSNR 
40 to 47.88 dB 
Security, Tamper detection and authentication, Robustness 
DCTDNA, Chaotic 
2.NCC 
0.92 to 0.98 

3.BER 
0.11 to 0.39 

Proposed Work 
1.PSNR 
42.89 to 59.89 
Robustness, Imperceptibility, Security 
DWTSVD, Arnold Transform, Chaotic Maps, Channel Coding 
2. NCC 
0.96 to 0.97 

3 MSE 
30.78 

4.BER 
0.56 

5.MSSIM 
0.48 
Determining image encryption from the dual encryption method, which will be the important factor of the scheme to extend this research in future research work, and the number of attacks tested on images increases in the future.