Combined Analysis of Visual Cryptography using SVD Technique and Frequency Domain Watermarking Technique

DOI : 10.17577/IJERTV4IS050361

Download Full-Text PDF Cite this Publication

Text Only Version

Combined Analysis of Visual Cryptography using SVD Technique and Frequency Domain Watermarking Technique

Madhusudhana B S

M.Tech Dept. of ECE, Dayananda Sagar College of engineering

Bangalore, India

Mrs. Sapna P J

Assistant professor, Dept. of ECE Dayananda Sagar college of Engineering, Bangalore, India

AbstractThe proposed method deals with the visual cryptography using a new combined technique called SVD technique with frequency domain watermarking technique. SVD technique and Frequency domain watermarking technique results in high PSNR, moderate the difficulty of the design and moderate time required for computation and compared to predictable VC techniques. In this paper Frequency Domain (FD) watermarking technique is used to prevent distorts attacks in the VC scheme. FD watermarking technique provides more authentications to secret image and it gives high robustness.

Keywords SVD, VC, Watermark, FD


    The visual cryptography was presented for binary images by Moni and Noar in 1994. Later, emerging in visual cryptography (VC) ,it is elaborate to gray scale images and color images. This paper involves segmenting binary images into transparent image; the original image is obtained by overlapping the transparent image in a specified manner. This technique requires a mathematical calculation process. Conventional technique was limited to the transmission of text documents and written documents which are overcome by new technology by sending gray scale images.

    SVD is an image compression technique applied to visual cryptography to generate secret shares. Watermarking ideology is a data protecting concept in which an image is used to hide the secret data. Depending on the identification of secret data the Watermarking technique is classified as 1) Blind 2) Semi-blind and 3) Non-blind. Here type used for to protect the data is Non-blind Watermark method. In the Frequency domain methodology discreet wavelet transforms(DWT), discreet wavelet transforms Discreet cosine transforms [6] and DWT DCT SVD techniques can be used. This paper implements the Frequency domain Non- blind watermarking technique. Combined techniques namely DWT DCT and SVD is used to give high depth of image hiding.


    In [1], Visual cryptography for grayscale images(GI) using optimization technique. Augmenting technique is achieved by taking the average of 4×4 blocks; this is applied to whole image to form an average image. The shares should be safeguard by putting in the host image using LSB watermark technique. One of the watermark ideology is a simple technique.

    In [2].VC is applied to grayscale image, here image is converted into bit plane. To binary plane visual cryptography is applied. Here binary one is represented by [1 1] or [1 0] or [0 1], binary zero is represented by [0 0]. The lengthof the shares generated are not equal. This method gives less PSNR with distorted re-constructed image image quality is less; this method fails to extract the input message with good PSNR.

    In [3], Author explains the concept of visual cryptography for images to get secure image. And author adds the concepts of VC to color images. This paper is concentrating on grayscale images and color images. Shares are constructed by considering the RGB color depth. The VSS(visual secret sharing) scheme is applied to expand the pixel and to maintain the good quality of reconstructed image.



    Partition image into 4×4 blocks

    Implementation of proposed method is as shown in figure 1,



    Secret image

    Visual Cryptography


    Secret image

    Watermark extraction

    Watermarking embedding

    Fig. 1.Block diagram of visual cryptography and watermark technique

    1. Share generation

      Secret image

      Partition image into 4×4 blocks

      Singular Value Decomposition technique

      Visual cryptography

      Shares generated Fig. 2.Block diagram of share generation

      Share generation includes the processing the secret image and partitioning this image and singular value decomposition (SVD) is applied to obtain mean and variance values to which VC is applied to obtain the shares. Here GI is considered and it is the secret image; this image has to be protected. To get the singular value decomposition, adjust the size of the image by appending the 0s so that image can be segregate by four . Segregate the data into 4×4 blocks and get decomposition of singular value

      Find the mean and standard deviation of 4×4 blocks. Replace the SVD values with the mean and standard deviation values and assign to share1 and share2. µ and Std.D(standard deviation) values are converted to binary form and assign to share3 and share4.

    2. Flowchart for share genertion


    Read image and adjust to divide by 4


    Convert the mean and standard deviation values to binary values and assign mean value to share3 and standard deviation value to share4

    Four Shares share1, share2, share3, share4 are generated


    Fig. 3 Flowchart of visual cryptography

  4. Frequency Domain (FD) watermark technique Watermark technique can be applied to images in time

    domain or in frequency domain. In time domain approach image is embedded directly by modifying the pixel values. Frequency watermarking technique can be achieved by using only DWT or DCT DWT or DCT, DWT and SVD transforms [4].

    Watermark technique involves watermark embedding and watermark extraction. The block diagram represented in figure 3shows that cover image is first adjusted to the size of the secret image the cover image and secret image is applied to watermark embedder, the output of which is watermarked image. This watermarked image is moved in a network. Received watermarked image is applied to watermark extractor to get back the secret image.

    Share1 extraction


    Apply SVD to 4X4 blocks of images

    DWT, DCT and SVD

    watermark technique

    Calculate mean and standard deviation for each block



    Cover image

    Replace the SVD values with mean and standard deviation

    Share3 DWT, DCT and SVD

    Replace mean and standard deviation values in modified image with 1 and 0 respectively in modified SVD image

    watermark technique


    Share3 extraction

    Share2 extraction

    DWT, DCT and SVD

    watermark technique


    Share4 extraction


    Append the first two rows to make 8 bits, assign 8 bits to share 1 and apply same procedure for share 2 generation

    DWT, DCT and SVD

    watermark technique

    Fig 4: Block diagram of watermark technique



    1. DCT

      DCT transform a signal from time domain to frequency domain. The equations for the transformations are as shown below.



      DCT is mainly employed in jpeg compression method [6]; it gives a DC coefficients and AC coefficients. Most of the information is carried by DC value which is in a position of (0, 0) in the matrix. Generally DCT is applied to 8×8 matrices. Here it is applied for the 2×2 matrix.

    2. DWT(Discrete wavelet transform.)

      DWT gives the relation between time and frequency domain of signal. Applying DWT to an image gives four different bands among which three are high frequency bands and one is low frequency band. Mot of the information is carried by low frequency band. So any operation/processing on image is carried by taking the low frequency band.





      Fig 5: Frequency bands results from the DWT technique

    3. SVD(singular value decomposion)

    SVD is compression technique, SVD of an image gives three matrices among these two matrices are ortho-normal matrices and one matrix is diagonal matrix. Diagonal elements are called as singular values. The equation for SVD is as

    Not all the singular values are used to reconstruct the original information; few singular values are enough to reconstruct the information, there by compression is achieved.



    Read cover image and adjust it to shares

    Read the shares(S)

    apply the DWT to cover image. Select LL band

    Apply DCT to LL band for 2×2 blocks and Collect the DC values

    Apply SVD to DC matrix and generate A1,B1 and D1 matrices

    Generate a matrix Dm=D1+S

    Apply SVD to Dm and generate A2,B2 and D2 matrices

    Modify DC values by replacing D1 by D2

    Apply IDCT to modified DC values

    Take IDWT to LL, HH, HL, LH bands

    Watermarked image


    Fig. 6. Flowchart for watermark embedding



    Watermarked image

    Apply the DWT to watermark image.

    Select LL band

    Apply DCT to LL band for 2×2 blocks and Collect the DC values and form a matrix Cr

    Apply SVD to the matrix Cr, get Ar,Br and Dr matrices

    Generate a new matrix E=A2*Dr*transpose (B2)



    Fig. 7. Flowchart for watermark extraction

    1. Steps involved in watermark extraction

      • Apply DWT of the watermarked image Sm get the 4 different frequency bands named as HH, LL, LH and HL bands.

      • Select LL band and Perform 2×2 DCT on LL band.

      • Collect the dc values and form a new matrix called Cr, which contains DC coefficients.

      • Perform SVD to matrix Cr results 3 matrices named as Ar, Br,

      • Generate a new matrix E=A2*Dr*transpose (B2)

      • Share reconstructed is given by S=E-D1.

  7. Results and Analysis

The experimental results are obtained by running in Matlab v8.1 and the results are shown below.

  1. Shares generation using Visual cryptography

    Original image

    SVD image





    Fig. 8. Share generation using SVD technique

  2. Watermark embedding





    Share1Water marked image

    Share2 Watermarked image

    Share3 Watermarke d image

    Share4 Waterm arked image

    Fig. 9. Embedding secret image on watermark image

    Share 1Watermarke d image

    Share2 Watermarked image

    Share3 Watermarke d image

    Share4Wate rmarked image





    Reconstructed image

  3. Watermark extraction

Fig. 10.extraction of shares and reconstructing original image

Propose method is applied for different images and resulting psnr is as shown in tabel I.

TABLE I: Reconstructed images PSNR





image)in dB













Boy. Jpg



TABLE II: Time elapsed for program completion


Here the combined analysis of visual cryptography and the watermarking technique is done. Here new technique is implemented where in SVD method is combined with VC to obtain good image quality with higher PSNR. This method is simpler compared to other techniques and takes less computation time. New approach DCT, SVD and DWT based Watermarking technique discussed gives good results in terms of image quality, noise resistance, PSNR, compared to conventional methods.


  1. Ayan Banergee and Shreya Banergee,A Robust Visual Vryptography for Photographic grayscale Images Using Bock optimization and Blind Invisible Watermarking International Journal of Computer Theory and Engineering, vol. 4, No. 2, April 2012.

  2. D. Tagaddos and A. Latif ,Visual cryptography for grayscale images using Bit plane, International journal of information Hiding and Multimedia Signal Processing., vol. 5, january 2014.

  3. Archana B. Dhole and Prof. Nitin J. janwe,An Implementation of Algorith in Visual Cryptography in Images Iinternational Journal of Scientific and Research Publications, vol. 3, March 2013.

  4. Mayank Awasthi and Himanshi Lodhi,Robust Image Watermarking based on Discret Wavelet Transform, Discret Cosine Transform and Singular Value Decomposion Advance in Electronic and Electric Engineering, vol. 3, pp. 971-976, 2013.

  5. Md Saiful Islam and Ui Pil Chang,a Digital Watermarking Algorith Based on DWT,DCT and SVD International Journal of Computer and Communication Engineering, vol. 3,no. 5, September 2014.

  6. Anitha S,Image Compression using Discret Cosone Transform and Discret Wavelt Transform International Journal Scientific and Engineering Research, vol. 2, November 2011

  7. Md. Maklachur Rahman,A DWT, DCT And SVD Based Watermark Technique To Protect The Image Privacy International Journal of Managing Public Sector Information and Communication Technologies, vol. 4, No. 2, June 2013.


Elapsed Time



27.66 S


26.54 S


28.66 S


26.5 S


28.193 S

Leave a Reply