1 Introduction

Information technology has eased the duplication, manipulation and distribution of digital data in recent times which has resulted in the demand for safe ownership of digital multimedia contents such as images, video and audio. The multimedia content-authentication and copyright protection has become an ongoing and constant requirement for protecting the media [6]. Recently, telemedicine applications play an important role in the development of the medical field. However, protect the transmission, storage and sharing of electronic patient record (EPR) data between two hospitals or via open channel are the most important issues in this field. Also, the digital imaging and communications in medicine (DICOM) is a basic criterion to communicate EPR data. A header is attached with the DICOM medical image files which contain important information about the patient, but it may be lost, attacked or disorder with other header file.

Digital watermarking is a technique which provides a best solution to these important issues [19]. The other advantages of the medical image watermarking are listed below [22]: 1) Storage space required for the image and the patient record is very voluminous. For small hospitals financial economy is an important decisive factor. Embedding the data in the corresponding images will save so much of the storage space. 2) Huge amount of bandwidth is required for the transmission of the image data for telemedicine purposes. The additional requirement of bandwidth for the transmission of the metadata can be avoided if the data is hidden in the image itself. Since the EPR and the image embedded into one, bandwidth for the transmission can be reduced in telemedicine applications. 3) Normally a patient does not like to expose his medical report to public; especially if the disease is of clandestine nature. The utmost confidentiality can be maintained by hiding the data in the image. When the patient data text and image are sent separately, if a tampering is done on the text or image, the after-effects may even cost a life due to wrong diagnosis. Most of the internet users are ignorant of the esoteric terms of watermarking techniques. When the patient data is hidden in medical image, such people will not try to tamper it. If tampered, it can be made out by using appropriate watermarking technique. Moreover, medical identity theft is a growing and dangerous crime as reported in various surveys [2, 25]. This demands development of secure medical data/image watermarking schemes.

The digital watermarking is a technique for inserting information into a cover massages (text, image, audio or video data) and later extracted or detected for variety of purposes including identification and authentication. Many wavelet based watermarking schemes were proposed for medical images [7, 10, 11, 18, 26, 35, 38, 40, 42]. Depending upon the type of data to be watermarked, the watermarking methods can be classified into four categories: text watermarking, image watermarking, audio watermarking, and video watermarking [32]. However, due to higher data embedding capacity of image covers, the present work focuses on watermarking using medical image as cover media. The image watermarking methods can be further classified as ‘spatial domain’ and ‘transform domain’ methods. Spatial domain methods are straight forward and computationally simple. LSB substitution, spread spectrum, and patchwork are the important spatial domain techniques. In spatial domain watermarking the watermark data is embedded directly by manipulating the pixel values, bit stream or code values of the host signal (cover media). However, spatial domain techniques are offer less robustness against attacks.

A spatial domain based digital image watermarking method is proposed by Lin et al. [12]. In this method, the host image is lossless and the watermark is robust to malicious attacks including geometric and non-geometric attacks. In the embedding process, the watermark logo is fused with noise bits first, and later XORed with the feature value of the image by 1/T rate forward error correction (FEC), where T is the times of data redundancy. During extraction process, the watermark bits are determined by majority voting. The experimental results have been shown that not only the image is lossless but also the proposed method can effectively resist the common malicious attacks. Also, the proposed method does not transfer the protected images to transform domain, and hence it can reduce the computation time of embedding and extraction process.

In the transform domain techniques the data is embedded by modulating the coefficients of a transform like discrete Fourier transform (DFT), discrete cosine transform (DCT), discrete wavelet transform (DWT) and singular value decomposition (SVD). The transform domain watermarking techniques are computationally complex but they provide greater robustness of watermarked data. Recently, the higher robustness of watermark has been achieved by using wavelet based watermarking are presented in [13–15]. The overall performance of the wavelet based watermarking technique depends greatly on embedding and extraction process. The main advantages of wavelet transform techniques for watermarking applications are: space frequency localization, multi-resolution representation, multi-scale analysis, adaptability and linear complexity. The wavelet based watermarking is also compatible with the new image standard JPEG 2000. Further, performance improvement of the watermarking methods using hybrid watermarking has been proposed by some researchers [1, 4, 5, 8, 9, 16, 20, 23, 28–31, 34, 36, 39]

Rosiyadi et al. [28] proposed a hybrid watermarking method based on DCT and SVD for the copyright protection. In this embedding process, DCT is applied on the host image using the zigzag space-filling curve (SFC) for the DCT coefficients and subsequently the SVD is applied on the DCT coefficients. Finally, the host image is modified by the left singular vectors and the singular values of the DCT coefficients to embed the watermark image. In this method, Genetic Algorithm (GA) based technique is used to find the optimization scaling factor of the watermark image. They have experimentally shown that the proposed method is robust against several kinds of attacks. The comparison between the method based on DCT and SVD using GA and the hybrid method based on DCT-SVD has been presented by Rosiyadi et al. in [29]. It is shown that the robustness of the extracted watermark and the visual quality of the watermarked image of the method using GA technique is better than the hybrid method. Horng et al. [31] proposed a blind watermarking method based on DCT, SVD and GA. It is shown that this method is robust and offers high imperceptibility against several known attacks.

Horng et al. [30] proposed an adaptive watermarking method based on DCT, SVD and GA. In this embedding process, the host image luminance masking is used and the mask of each sub-band area is transformed into frequency domain. Subsequently, the watermark image is embedded by modifying the singular values of DCT-transformed host image with singular values of mask coefficients of host image and the control parameter of DCT-transformed watermark image using GA. It is shown that this method is robust against several known attacks.

Consequently, many other image watermarking techniques combining three transform methods have been proposed [1, 4, 5, 8, 9, 16, 20, 23, 34, 36, 39]. For a detailed description on these combined approaches, interested readers may directly refer to them.

Singh et al. [32] proposed a robust hybrid watermarking technique using fusion of DWT, DCT, and SVD instead of applying DWT, DCT and SVD individually or combination of DWT-SVD / DCT-SVD. The suggested technique initially decomposes the host image into first level DWT followed by transformation of Low frequency band (LL) and watermark image using DCT and SVD. Then the S vector of watermark image is embedded in the S component of the host image and the watermarked image is generated by inverse SVD on modified S vector and original U, V vectors followed by inverse DCT and inverse DWT. The watermark is extracted using an extraction algorithm. The proposed method has been extensively tested and analyzed against known attacks.

Terzija et al. [37] have proposed a method for improving efficiency and robustness of the watermarks using three different error correction codes namely, (15,7)-BCH, (7,4)-Hamming Code and (15-7)-Reed-Solomon code. These codes are applied to the ASCII representation of the text which is used as watermark. The watermark is embedded into the original cover image by first decomposing the cover up to second level using discrete wavelet transform (DWT) with the pyramidal structure and then the watermark is added to the largest DWT coefficients that represent high and middle frequencies of the cover image. It is shown that Reed-Solomon code performs better due to its excellent ability to correct errors, however, the ECCs considered are not able to deal with bit error rates (BER) greater than 10–20 %.

In this paper, we are using Singh method to embed the image watermark and Terzija method to embed text watermark. Results are obtained by varying the gain factor, size of watermark, and medical image modalities. Experimental results are provided to illustrate that the proposed method is able to withstand a known attacks.

The following observation are apparent:

  1. 1)

    Capacity of embedding multiple watermark: The method proposed in [1, 4, 5, 8, 9, 16, 20, 23, 34, 36, 39] has been embedded only one watermark. However, in the proposed method multiple watermarks (text and image) are embedded simultaneously, which provides extra level of security with acceptable performance in terms of robustness and imperceptibility. For identity authentication purposes, multiple watermarks are embedded instead of single watermark into the same medical image / multimedia objects simultaneously, which offer superior performance in telemedicine and tele-diagnosis applications. The simultaneous multiple watermarking embedding method has fewer constraints than the other multiple watermarks embedding methods [33].

  2. 2)

    Improved robustness performance: In Table 5, The robustness of proposed method is compared with other reported techniques [29, 34, 36] and it is found that the proposed method offers superior performance

  3. 3)

    Enhance the Security: Security of the medical text watermark may be enhanced by using simple encryption method to save execution time. For telediagnosis, the encryption and decryption speed has become an important factor if the situation demands.

  4. 4)

    Reduced storage and bandwidth requirements: The medical image files / electronic patient record (EPR) contain important patient data. Further, in order to conserve the transmission bandwidth or storage space the patient’s details may be embedded inside the medical image.

Therefore the proposed method may find potential application in prevention of patient identity theft in telemedicine applications

2 Theoretical background

The proposed multiple watermarking method based on DWT, DCT and SVD. The watermark image will be discrete Cosine transformed at first. The DCT information of the watermark image contains low frequency information and as long as such information is not lost or lost a little, the watermarking image can be extracted well. One of important mathematical properties of SVD is that slight variations of singular values do not affect the visual perception of the cover image, which motivates the watermark embedding procedure to achieve better quality of the watermarked image and robustness of the extracted watermark. Hence, a brief description of these concepts is included in the given below sections.

2.1 Discrete wavelet transform (DWT)

The DWT has received considerable attention in various signal processing applications, including image watermarking. The main idea behind DWT results from multi-resolution analysis, which involves decomposition of an image in frequency channels of constant bandwidth on a logarithmic scale. It has advantages such as similarity of data structure with respect to the resolution and available decomposition at any level [11]. DWT separates an image into a set of four non- overlapping multi-resolution sub bands denoted as lower resolution approximation image (LL) as well as horizontal (HL), vertical (LH) and diagonal (HH) detail components. The process can then be repeated to compute multiple scale wavelet decomposition. Since human eyes are much more sensitive to the low-frequency part (LL sub-band), the watermark can be embedded into the other three sub-bands (HL, LH and HH sub-band) to maintain better image quality. It is evident that the energy of an image is concentrated in the high decomposition levels corresponding to the perceptually significant low frequency coefficients; the low decomposition levels accumulate a minor energy proportion, thus being vulnerable to image alterations. Therefore, watermarks containing crucial medical information such as doctor’s reference, patient identification code, image codes etc. requiring great robustness are embedded in higher level sub-bands [3].

2.2 Discrete cosine transform (DCT)

The discrete cosine transform (DCT) works by separating image into parts of different frequencies, low, high and middle frequency coefficients [32], makes it much easier to embed the watermark information into middle frequency band that give additional resistance to the lossy compression techniques, while avoiding significant modification of the cover image. The DCT has a very good energy compaction property. For an input image, I, of size N x N the DCT coefficients for the transformed output image, D, are computed according to Eq. (1). I (x,y) is the intensity of the pixel in row x and column y of the image, and D (i, j) is the DCT coefficient in row i and column j of the DCT matrix.

$$ \begin{array}{l}D\left(i,j\right)=\frac{1}{\sqrt{2N}}C(i)C(j){{\displaystyle \sum}}_{x=0}^{N-1}{{\displaystyle \sum}}_{y=0}^{N-1}I\left(x,y\right) cos\frac{\left(2x+1\right)i\pi }{2N} cos\frac{\left(2y+1\right)i\pi }{2N}\hfill \\ {}\begin{array}{lll}\mathrm{C}\left(\mathrm{i}\right),\mathrm{C}\left(\mathrm{j}\right)=\frac{1}{\sqrt{\mathrm{N}}}\ \mathrm{f}\mathrm{o}\mathrm{r}\ \mathrm{i},\kern0.5em \mathrm{j}=0\hfill & \mathrm{and}\hfill & \mathrm{C}\left(\mathrm{i}\right),\mathrm{C}\left(\mathrm{j}\right)=\sqrt{\frac{2}{\mathrm{N}}}\ \mathrm{f}\mathrm{o}\mathrm{r}\ \mathrm{i},\mathrm{j}=1,2,\dots \dots .\mathrm{N}-1\hfill \end{array}\hfill \end{array} $$
(1)

2.3 Singular value decomposition (SVD)

The singular value decomposition of a rectangular matrix R m is a decomposition of the form

$$ {\mathrm{R}}_{\mathrm{m}}={\mathrm{USV}}^{\mathrm{T}} $$
(2)

Where R m is an M × N matrix, U and V are orthonormal matrices, S is a diagonal matrix comprised of singular values of R m . The singular values s1≥s2≥………≥sn≥0 appear in descending order along the main diagonal of S. These singular value are obtained by taking the square root of the Eigen values of R m R T m and R T m R m . The singular values are unique; however the matrices U and V are not unique [32].

2.4 Encryption and decryption process of EPR data

For providing additional security, text watermark (EPR data) may be encrypted before watermarking. However, the delay encountered during embedding and extraction of the watermark is also an important factor in telemedicine applications. Therefore, watermarking methods using encryption techniques should be simple to save execution time [41]. The text watermark in the proposed method is encrypted using the equation

$$ \mathrm{Encrypted}\kern0.5em \mathrm{text}\kern0.5em \mathrm{watermark}=\left(\mathrm{text}\kern0.5em {\mathrm{watermark}}^{\mathrm{r}}\right)-\mathrm{d} $$
(3)

where r and d are constants. Here, r can have a value in the range 1.000 to 1.143 and d can be between 0.0 and 10.0. The first level of security lies in this encryption process [21].

The extracted encrypted text is decrypted at the receiving end using the relation

$$ \mathrm{Decrypted}\kern0.5em \mathrm{text}\kern0.5em \mathrm{watermark}={\left(\mathrm{Encrypted}\kern0.5em \mathrm{text}\kern0.5em \mathrm{watermark}+\mathrm{d}\right)}^{\frac{1}{\mathrm{r}}} $$
(4)

3 Performance measures

The performance of the watermarking algorithm can be evaluated on the basis of its robustness and imperceptibility. A larger Peak Signal to Noise Ratio (PSNR) indicates that the watermarked image more closely resembles the original image meaning that the watermark is more imperceptible. Generally, watermarked image with PSNR value greater than 27 is acceptable [32]. The PSNR is defined as

$$ \mathrm{PSNR}=10 \log \frac{{\left(\mathrm{Bmax}\right)}^2}{\mathrm{MSE}} $$
(5)

Where Bmax is maximum pixel value of the image, the Mean Square Error (MSE) is defined as

$$ \mathrm{M}\mathrm{S}\mathrm{E}=\frac{1}{\mathrm{X}\times \mathrm{Y}}{\displaystyle {\sum}_{\mathrm{i}=1}^{\mathrm{X}}{\displaystyle {\sum}_{\mathrm{j}=1}^{\mathrm{Y}}{\left({\mathrm{I}}_{\mathrm{i}\mathrm{j}}-{\mathrm{W}}_{\mathrm{i}\mathrm{j}}\right)}^2}} $$
(6)

where Iij is a pixel of the original image of size X × Y and Wij is a pixel of the watermarked image of size X × Y. The robustness of the algorithm determined in term of correlation factor. The similarity and differences between original ‘watermark and extracted watermark is measured by the Normalized Correlation (NC). Its value is generally 0 to 1. Ideally it should be 1 but the value 0.7 is acceptable [32].

$$ \mathrm{N}\mathrm{C}={\displaystyle {\sum}_{\mathrm{i}=1}^{\mathrm{X}}}{\displaystyle {\sum}_{\mathrm{j}=1}^{\mathrm{Y}}}\left({\mathrm{W}}_{\mathrm{originalij}}\times {\mathrm{W}}_{\mathrm{recoveredij}}\right)/{\displaystyle {\sum}_{\mathrm{i}=1}^{\mathrm{X}}}{\displaystyle {\sum}_{\mathrm{j}=1}^{\mathrm{Y}}}{\mathrm{W}}_{\mathrm{originalij}}^2 $$
(7)

where Woriginalij is a pixel of the original watermark of size X × Y and Wrecoveredij is a pixel of the recovered watermark of size X × Y.

The bit error rate (BER) [33] is defined as ratio between number of incorrectly decoded bits and total number of bits. It is suitable for random binary sequence watermark. Ideally it should be zero.

$$ \mathrm{B}\mathrm{E}\mathrm{R} = \left(\mathrm{Number}\ \mathrm{of}\ \mathrm{incorrectly}\ \mathrm{decoded}\ \mathrm{bits}\right)/\left(\mathrm{Total}\ \mathrm{number}\ \mathrm{of}\ \mathrm{bits}\right) $$
(8)

4 Proposed method

The method proposed is a combined DWT, DCT and SVD based process. The proposed method increases the robustness without significant degradation of image quality against the signal processing attacks. Figure 1a and b illustrates the multiple watermark embedding and extraction process respectively. The algorithmic steps are discussed below:

Fig. 1
figure 1

Multiple watermarks a embedding process and b extraction process

Full size image

4.1 Embedding algorithm for image watermark

start:

STEP 1: Variable Declaration

Barbara Image: cover image

Medical Image (Thorax): watermark image

C_w: read the cover image

W_w: read the watermark image

α: gain factor

DWT, DCT and SVD: Transform Domain Techniques

Wavelet filters: Haar

LL c , LH c , HL c , and HH c : First level DWT coefficients for cover image

D: DCT coefficients of watermark image

D 1 c : DCT coefficients matrix for HH c

U c   and V T c : orthonormal matrices for D 1 c

S c : diagonal matrix for D 1 c

U w   and V T w : orthonormal matrices for D

S w : diagonal matrix for D

W k w : modified value of S c

U ww   and V T ww : orthonormal matrices for W k w

S ww : diagonal matrix for W k w

W modi : Modified DWT coefficient

W idct : InverseDCT coefficients matrix

W d : Watermarked Image

STEP 2: Read the Images

C_w ← MRI.bmp (Cover image of size 512*512)

W_w ← Thorax.bmp (Watermark image of size 256*256)

STEP 3: Perform DWT on Cover and DCT on Watermark image

Apply first level DWT on cover image

[LL c , LH c , HL c , HH c ] ← DWT(C_w, wavelet filter);

D = DCT(W_w);

STEP 4: Choice of subands in Cover and obtain the DCT coefficients for the same

//Choose sub-band LL c from cover image

if (DCT on LL c ) then

D 1 c  ← DCT (LL c );

endif;

STEP 5: Compute the singular values of DCT coefficients for Cover and Watermark image

if (SVD on D 1 c ) then

U c S c V T c  ← SVD(D 1 c )

endif;

if (SVD on D) then

U w S w V T w  ← SVD(D)

endif;

STEP 6: Watermark Embedding

for α ← 0.1:0.9

S c  + αS w  = W k w ;

end;

STEP 7: Compute the singular values for W k w and obtain the modified DWT coefficients

if (SVD on W k w ) then

U ww S ww V T ww  ← SVD(W k w )

endif;

//modified DWT coefficient

W modi  ← U c S ww V T c

Step 8: Obtain the Watermarked Image.

W idct  ← inverse(W modi );

//Apply InverseDWT to LL c , LH c , HL c  and HH c with modified coefficient

W d  ← InverseDWT (W idct , LH c , HL c , HH c wavelet filter);

end:

4.2 Extraction algorithm for image watermark

start:

STEP 1: Variable Declaration

α scale factor

LL c , LH c , HL c , HH c : sub-bands for watermarked image

D * w : DCT coefficients matrix for HH c

U * w   and V * T w : orthonormal matrices for D * w

S * w : diagonal matrix for D * w

S * k: modified values

U * 1 w   and V * 1T w : orthonormal matrices for S * k

S * 1 w : diagonal matrix for S * k

I * cc : modified DWT coefficients

W EW : Extracted watermark image

STEP 2: Perform DWT on Watermarked image (possibly distorted)

[LL c , LH c , HL c , HH c ] ← DWT (W d , wavelet filter);

STEP 3: Obtain the DCT coefficients for HH c

if (DCT on LL c ) then

D * w  ← DCT (LL c );

endif;

STEP 4: Compute the singular values for D * w

U * w S * w V * T w  ← SVD(D * w )

end;

STEP 5: Perform the operation and then apply SVD

for α =0.1:0.9

\( {S}^{*k}=\frac{S_w^{*}-{S}_c}{\alpha } \)

end;

U * 1 w S * 1 w V * 1T w  ← SVD(S * k)

STEP 6: Compute modified DWT coefficients

I * cc  ← U w S * 1 w V T w

STEP 7: Extract the watermark image.

W EW  ← InverseDCT (I * cc );

end:

4.3 Embedding algorithm for text watermark

start:

STEP 1: Variable Declaration

Medical Image(MRI): cover image

Leena: watermark image

C_w: read the cover image

W_w: read the text watermark

α: scale factor

DWT: discrete wavelet transforms

Wavelet filters: Haar

LL c , HL c , LH c  and HH c : First level DWT coefficients for cover image

LL c1, HL c1, LH c1 and HH c1: Second level DWT coefficients for cover image

STEP 2: Read the Images

M_w ← MRI.bmp (Cover image of size 512*512)

STEP 3: Perform DWT on Cover image

//Apply second level DWT on cover image

[LL c , HL c , LH c  and HH c ] ← DWT (M_w, wavelet filter);

[LL c1, HL c1, LH c1 and HH c1] ← DWT (HH c , wavelet filter);

STEP 4: Encrypt the watermark text by using Eq. ( 3 )

STEP 5: Convert encrypted watermarking text to Binary bits

// converting text watermark into binary bits

Wtxt ← binary(Text Watermark);

STEP 6: Replace ‘(0,1)’ by ‘(−1,1)’ in the watermarking bits

// bit stream is transformed into a sequence w(1) w(2)....w(L) by replacing the 0 by −1 and 1 by 1, L is the length of string

− 1 ← 0 and 1 ← 1;

STEP 8: Embedding the text watermark

// text watermark is embeds into HH c1 sub-band

for ∝ ← 0.01 : 0.1

f '(x, y) = f(x, y)(1 + ∝ × Wb); f(x, y) and f '(x, y) is DWT coefficients before and after embedding process

end;

STEP 9: Obtain the Watermarked Image W d

//Apply Inverse DWT to LLc, HLc, LHc and HHc with modified and unmodified DWT coefficients

\( \begin{array}{l}{W}_{mg}= inverse\kern0.5em DWT\left(L{L}_{c1},\kern0.5em H{L}_{c1},\kern0.5em L{H}_{c1}\kern0.5em and\kern0.5em H{H}_{c1},\kern0.5em wavelet\kern0.5em filter\right);\hfill \\ {}{W}_d\leftarrow inverse\kern0.5em DWT\left(L{L}_c,\kern0.5em H{L}_c,\kern0.5em L{H}_c\kern0.5em and\kern0.5em {W}_{mg},\kern0.5em wavelet\kern0.5em filter\right);\hfill \end{array} \)

end:

4.4 Extraction algorithm for text watermark

In the watermark extraction procedure, both the received image and the cover image are decomposed into the two levels. It is assumed that the cover image is available for extraction process

start:

STEP 1: Variable Declaration

Medical Image (MRI): cover image

Leena: watermark image

C_w: read the cover image

α: scale factor

DWT: discrete wavelet transforms

Wavelet filters: Haar

LL c , HL c , LH c  and HH c : First level DWT coefficients for cover image

LL c1, HL c1, LH c1 and HH c1: Second level DWT coefficients for cover image

STEP 2: Perform DWT on Watermarked image (possibly distorted)

// original image is also available for extraction process

[LL c , HL c , LH c  and HH c , wavelet filter] ← DWT (W d , wavelet filter);

STEP 3: Watermark extraction

\( {W}_rb=\frac{\left({f}_r^{\prime}\left(x,y\right)-f\left(x,y\right)\right.}{\alpha f\left(x,y\right)} \); f r  ′ (x, y) are the DWT coefficients of the received image.

//finally extracted watermark taken as sign (either positive or negative)

W e b ← positive or negative sign(W r b);

STEP 4: Convert the watermark bits into text to get the characters

STEP 5: Decrypt the characters by using Eq. (4) to get the original watermark

end:

5 Experimental results and analysis

The performance of the proposed watermarking method by applying encryption on patient data before embedding the text watermark has been investigated. For testing the robustness and quality of the watermarked image MATLAB is used. In the proposed method cover image of size 512 × 512 [17], the image watermark of size 256 × 256 and the text watermark of size 50 characters are used for testing. The robustness of the image and text watermarks is evaluated by determining NC and BER respectively. The quality of the watermarked image is evaluated by PSNR. It is quite apparent that size of the watermark affects quality of the watermarked image. The size of the watermark is sum total of bits occupied by all watermarks in the case of multiple watermarking. However, degradation in quality of the watermarked image will not be observable if the size of watermark (total size in case of multiple watermarking) is small.

The image watermark (Thorax image) embedding method is based on DWT, DCT and SVD. In order to enhance the security of the text watermark, encryption is applied to the ASCII representation of the text watermark before embedding. It is found that larger gain factor results in stronger robustness of the extracted watermark whereas smaller gain factor provides better PSNR values between original and watermarked medical images. However, overall performance of the proposed method highly depends on the size of the watermarks, gain factor and the noise variation.

Figure 2a shows the CT Scan cover image and Fig. 2b–d show watermarked images at different gain factors 0.05, 0.5 and 1.0 respectively. Figure 3a shows the original image watermark (Thorax image). The text watermark is the patient data as shown in Fig. 3b. In the experiment, values of PSNR, NC and BER are illustrated in Table 1 to Table 3 for varying gain factor (α) in the range of 0.01 to 0.5. Furthermore, we have tested the robustness of the image watermark with checkmark benchmarking software (http://watermarking.unige.ch/Checkmark/) [27] which is shown in Table 4.

Fig. 2
figure 2

Original and watermarked CT scan images a original image and watermarked images with gain factor; b 0.05; c 0.5 and d 1.0

Full size image
Fig. 3
figure 3

Original a image and b text watermark

Full size image
Table 1 Performance of proposed method at different gain factor
Full size table

In Table 1, performance of the proposed method against different size of watermark has been evaluated without any noise attack. With the encryption, maximum PSNR value is 35.84 dB and BER = 0 against maximum size of watermark at gain factor = 0.01. Here, the NC value is 0.9802. However, the maximum NC value is 0.9992 at gain factor = 0.1. Table 2 shows the performance of the proposed watermarking method against different attacks. With encryption, the highest BER value of 0.96 has been obtained against JPEG Compression with quality factor = 10. Figure 4 shows the performance of the proposed method against known attacks.

Table 2 Performance of proposed method against different attacks at gain = 0.05
Full size table
Fig. 4
figure 4

Performance of the proposed method against different signal processing attacks

Full size image

Table 3 shows the effect of cover image as proposed method was tested for other types of cover images like Brain CT Scan, ultrasound, Barbara and Lena images. The highest PSNR were obtained with ultrasound image at α = 0.05. Here, the NC and BER value is 0.9983 and 0.6 respectively. However, the maximum NC and BER were obtained with Lena and MRI image respectively. Figure 5 shows the effect of the cover image on the parameter PSNR, BER and NC values.

Table 3 Effect of cover image at gain = 0.05
Full size table
Fig. 5
figure 5

Performance of the proposed method against different cover images

Full size image

Table 4 shows the performance of the proposed method against checkmark attacks. The maximum NC values of the extracted watermark under different attacks is shown at α = 0.09. In this table, the maximum NC value of 0.9774 has been obtained by the proposed method against projective attack. However, the minimum NC value is 0.6189 against rows and columns removal attack. Here, all the NC values have been accepted except the rows and columns removal, trimmed mean and mid point attack which are less than 0.7. Figure 6 shows the robustness performance of the proposed method against checkmark attacks.

Table 4 Performance measure for checkmark attacks at gain = 0.09
Full size table
Fig. 6
figure 6

Performance of the proposed method against checkmark attacks

Full size image

Table 5 shows the performance comparison of the proposed method with other reported techniques. In this table, the NC value with proposed method has been obtained as 0.9994, 0.9752 and 0.6565 against JPEG, Median Filtering, and Gaussian Noise (Var-0.5) attacks respectively. However, the NC value obtained with Rosiyadi et al. [29] method is −0.1863, 0.4585 and 0.5012 against the same attacks respectively. The NC value with Singh and Tayal method [34] has been obtained as 0.9956, 0.8893, 0.7809 and 0.9636 against Gaussian LPF, Gaussian Noise (Var-0.01), Salt & Pepper Noise (Density = 0.08) and Histogram attacks respectively. However, the NC value obtained with proposed method is 0.99560.9754, 0.8856 and 0.9208 against the same attacks respectively. The NC value with Srivastav et al. [36] has been obtained as 0.6019, 0.632 and 0.9123 against Median Filtering, Gaussian Noise (Var-0.01) and Histogram attacks respectively. However, the NC value obtained with proposed method is 0.9752, 0.9754 and 0.9208 against the same attacks respectively.

Table 5 Performance comparison results under NC value
Full size table

Overall, the performance of the proposed method is better than the other reported technique [29, 34, 36] in terms of robustness, capacity and security. Finally, quality of the watermarked image has been measured by the subjective method [24]. Humans are involved to check and vote for the quality of the watermarked data. Table 6 reports their combined suggestion. It may be observed that the reported visual quality of the watermarked images is acceptable for diagnosis at all chosen gain factors except the gain factor (α) = 1.0 and 5.0, which indicate the poor quality. It may be concluded from the subjective measure test that smaller gain factor provides acceptable quality of the watermarked image for diagnosis.

Table 6 Subjective measure of the watermarked image quality at different gain factor
Full size table

6 Conclusions

In this paper, a hybrid image-watermarking technique based on DWT, DCT and SVD has been presented, where the watermark is embedded on the singular values of the cover image DWT sub bands. The main properties of this work can be identified as follows:

  1. 1)

    Proposed algorithm combines the advantages and removes the disadvantages of these three most popular transforms namely DWT, DCT and SVD. These are the novel techniques used for watermarking so their fusion makes a very attractive watermarking technique. Due to its excellent spatio-frequency localization properties, the DWT is very suitable to identify areas in the cover image where a watermark can be imperceptibly embedded. The DCT information of the watermark image contains low frequency information and as long as such information is not lost or lost a little, the watermarking image can be extracted well. One of attractive mathematical properties of SVD is that slight variations of singular values do not affect the visual perception of the cover image, which motivates the watermark embedding procedure to achieve better performance in terms of imperceptibility, robustness and capacity as compared to DWT, DCT and SVD applied individually or combination of DWT-SVD or DCT-SVD.

  2. 2)

    For identity authentication purposes, multiple watermarks have been embedded instead of single watermark into the same medical image / multimedia objects simultaneously, which offer superior performance in telemedicine and tele-diagnosis applications.

  3. 3)

    In the proposed method, we have embedded two watermarks simultaneously. This technique has fewer constraints than the other two dual watermarks embedding methods.

  4. 4)

    Security of the text watermark is enhanced by using encryption method. However, encrypting EPR data before watermarking has become unavoidable, but the delay encountered during embedding and extraction of the watermark is also an important factor in telemedicine applications. Therefore, watermark constitution by using encryption methods should be simple to save execution time. For telediagnosis, the speed has become an important factor if the situation demands. The encryption /decryption method used in the proposed technique is very simple.

Overall, the proposed method is better than the other reported technique in terms of robustness and security. This may provide a potential solution to existing telemedicine security problem of patient identity theft.

The inclusions of many techniques were combined to improve the robustness of the watermarks and the quality of the watermarked image which is the prime objective of the research. However, it may have increased the computational complexity to some extent which needs to be investigated separately. We also need to investigate approaches that will simultaneously improve the performance such as robustness, imperceptibility, security and capacity.

We would like to further improve the performance, which will be reported in future communication.