A performance Evaluation of Transform Domain Algorithm in Watermarking Based on Different Levels of Sub-Bands of Discrete Wavelet Transform

ABSTRACT


Introduction
With the new progress in the internet technology, it makes digital contents to be accessible by various users. However, this technology has advantages such as digital content can be access easily with less time, furthermore, digitalization also has disadvantages like illegal copy, distribution, reproduction and manipulation on digital contents without the knowledge of the owners [14]. These problems can be solved by applying watermark. Digital watermarking is a powerful scheme applied to covered data in order to protect illegal copy, reproduction and security issues of digital content like image, audio and text. Digital watermarking is done by applying strong, specific and applicable techniques which play a vital role in digital watermarking [1], [7]. Recently, digital watermarking has showed one of the crucial research aspects in the area of signal processing and information security. The aim is to insert un-removable digital watermarking into original content and also the same watermark can be extracted to solve copyright protection. Therefore, the inserted digital watermark must have the following characteristic; Imperceptibility and robustness [16]. Maximum of information; an embedded watermark data should have capacity to contain largest watermark information [10]. However, maximum of information indicates the amount of information that can be embedded into multimedia contents, capable of hiding watermarks that can be detected for the authentication and copyright protection based on condition of imperceptibility and robustness. However, embedding large watermark information it can be more difficult work in digital image watermarking. Maximum of information is the number of inserted bits into pixel of the digital image.

Related work
Several studies have been carried out in the field of watermarking. Some of them are as follow: Kaur et..al., developed a digital watermarking algorithm that based on DWT, FFT, Arnold transform and SVD. Watermarking embedding process is done by applying cover image on FFT, then applied DWT on image to decompose it into four sub-bands and then, applied SVD. The developed method implemented in MATLAB. PSNR is applied for imperceptibility and the obtained results proved that the developed technique is imperceptible [6].
Laura et al. proposed a digital image watermarking technique based on DWT, Chirp Z-transform, entropy and SVD. These techniques embed the digital watermark in colored image. The obtained results of the developed algorithm showed that is robustness and imperceptiblity against various signal operation attack like compression and histogram equalization [4].
Gunjal& Mali [2] proposed a watermarking scheme. This scheme is developed by using Java.net and MATLAB. The proposed scheme is based on DWT-Fast wavelet Hadamard transform. (DWT-FWHT-SVD) and DWT domains. However, orthogonal wavelet and symmetric is adopted for DWT decomposition and also Fiboacci-Lucast transform together with affine transforms that are applied for inserting security. The experimental result shows the good performance in DWT-FWHT-SVD domain than DWT domain in terms of security and robustness [2].
Rassem et al., described digital watermarked algorithms based on DWT -SVD and Redundant DWT -SVD, RDWT-SVD. These algorithms are to evaluate their performance based on robustness and imperceptibility. The obtained results proved that the proposed techniques are both robust against various attacks, but DWT-SVD it resists more attacks than RDWT-SVD such as scaling and filtering, JPEG compression [11].
Theakur, proposed digital watermarking algorithms based on the DWT, FFT and SVD for imperceptibility and robustness on digital image watermarking. The proposed method is implemented in MATLAB. The embedding process of this technique is done in such a way that, watermark signal can be embedded without affecting the quality of the original image. The proposed algorithm showed that is robust and imperceptible based on the experimental results [15].
Saxena et al, developed a watermarking technique based on DWT, DCT and SVD with semi-blind algorithm has been applied. The performance evaluation of the proposed technique is done by performing PSNR and CC for both imperceptibility and robustness. Various attacks have been applied in MATLAB to measure the robustness of the proposed method such as rotation, cropping, translation and noise. However, the obtained result showed that the propose algorithms is robust against several attacks and also imperceptible [13].
Pandurangan and logashanmugamproposed watermarking techniques by applying DWT-DCT with SVD transform on digital image. In inserting process, the DCT is carried out on the original image by using zigzag sequence, and then represent the DCT coefficient into four quadrants. Furthermore, SVD is used to each quadrant. The experimental result showed that there is good robustness and transparency on the proposed algorithm [9].
Savakar&Ghuli, proposed a watermarking technique m based on DWT-FFT on digital image. DWT is carried out on original image for watermarking embedding process. Then, FFT is applied to convert input signal from spatial domain format to a frequency domain format. The evaluation of the proposed techniques is done based on fragility and fidelity, MATLAB 2009a is used as programming tool for simulations. The experimental result demonstrated that, the fragility with use various attacks [12].
Patel &Nandurbarkar, developed a watermarking technique based on DCT-DWT-SVD, initially watermarking is based on SVD of DC coefficient by implementing a second-level of DWT decomposition and the next method is based on SVD of DCT values of the second-level DWT composition of original image. All these processes are implemented in MATLAB. The experimental results explained that both the two approaches have shown robustness and imperceptiblity against different signal processing [8].
Kang &Iwamura, proposed digital image watermarking technique using various frequency coefficients that derived from DCT values of the second level DWT transform. Bose-Chaudhuri-Hocquen(BCH) code is used for watermark and BCH code is applied for synchronous data. The proposed technique is used for statistical analysis to choose the best position in the frequency domain for watermark embeds [5].
original image is applied to several watermark attacks. The obtained results demonstrated that the developed scheme has shown robustness [3].

Proposed method
The proposed method consists of inserting and extracting watermark on digital on digital contents. Digital watermarking algorithm applied on this research is DWT-SVD based on resolution of different sub-bands level that is LL, LH, HL and HH on image to protect multimedia content from illegal copy, manipulation, reproduction and distribution of digital content over the internet. The digital images used for inserting watermark are; the host image used is House and the watermark applied is signature. In this research, the scheme developed by applied MATLAB 2015. The two grey images used in this method are House for original image while Signature for watermark image.

Process of Embedding image watermarking based on LL
The DWT-SVD digital watermark embedding technique explained below: 1. Apply Discrete Wavelet Transform in order to decompose the original image B into LL of one sub-band. As shown in the following equation.
[LL, LH, HL, HH] = DWT 2 (B) (3.1) 2. Apply SVD to LL of one level sub-band image K as given in this equation (3.2) 3. Apply SVD to the watermark image C as given in this equation The decomposed cover image that used SVD watermark with scaling factor. as given below The orthogonal matrix of the cover image is added along with SVD S 2 to that is watermarked.
The equation is given below LL2 = U * S 2 *V'

Performance analysis
This chapter evaluates the performance of the described technique, experimental result and lastly observations. This scheme is implemented in MATLAB 2015. The algorithm proposed is based on digital Image watermarking. Therefore, the original image used on this technique is House image while watermark data is signature. After the watermarked data the experimental result showed that, between the host images, and watermarked are identical. The PSNR is applied on watermarked image Therefore the PSNR is measured for all the four level of sub band of DWT techniques. That is, LL, LH, HL and HH the proposed scheme showed that PSNR has the higher values it means that the proposed method is imperceptible. Many attacks were applied on image to evaluate the robustness of the proposed method. NCC is used to evaluate robustness of digital image.

Experimental Result
The described scheme is analyzed by using robustness and imperceptibility. The imperceptibility is to show the qualities of the watermarked image. That is, the present of inserted watermark image could not make any changes on the originalimage. The robustness is used to evaluate the resistance of digital image watermark against various attacks and signal operations on images like salt and pepper, Gaussian noise, rotation, cropping,and scaling.

Imperceptibility Testing for Image
To evaluate imperceptibility of digital image watermarking, PSNR is applied on digital images

Robustness Testing for Image
To evaluate robustness of digital watermarking, NCC is used. The NCC is applied on digital image. In the field of watermarking the range of NCC is between 1.0 and -1.0. If the numbers of NCC is exactly one. It means there is a perfect identity between the original watermarked image and recovered watermarks