# INTRODUCTION

ow days the rapid growth of the technology is a cause for the requirement of the protection of copyright information from unauthorized access. Hence an advanced authentication method is essential to the data security. Digital watermarking is one such method that offers data security. At present various kinds of digital watermarking methods are in use to protect the information from unauthorized access. Any security method come under either spatial [1] like LSB or frequency domain [2] like Discrete Wavelet Transform (DWT) [5], Discrete Cosine Transform (DCT) [3], Discrete Fourier Transform (DFT) [6] and the combination of them [7,8] as well as any method can come under either blind or non blind watermarking. In blind watermarking the cover image is not used to decipher the watermark [3]. The cover image is used to decipher the watermark in the non-blind watermarking [4]. This paper proposes a non blind watermarking method based on the frequency domain. This method uses the transformations to improve the robustness of the system. The necessary features are perceptual quality and robustness to determine the quality of watermarking scheme.

In the frequency domain, DWT decomposes an image into multi-resolution components i.e. horizontal, vertical and diagonal [5] and DCT segregate each block into three frequency sub-bands: low, mid and high [3]. DFT requires an input image that is discrete. Such inputs are often created by sampling a continuous function. But, in this correspondence, DWT-DCT combination [7] and DCT coefficients of watermark are used in the proposed method. It improves all security factors in the data transfer. The strength of this paper is, analysis of the proposed scheme based on the standard metrics such as PSNR and NC and it sustains the general attacks like Gaussian noise, salt & pepper noise, Poisson noise, Gaussian blur, sharpening and common image processing operations like cropping, JPEG compression. The rest of the paper is organized in the following way. Section 2 consists of proposed method. Section 3 holds the description of performance metrics. Results and analysis are illustrated in Section 4.


# II.


# PROPOSED METHOD

The main aim of the proposed method is to improve the quality of watermarked image and robustness of the watermark. This approach consists of two major processes.

? Applying DWT and DCT on RGB cover image to get transformed image. ? Embed DCT coefficients of watermark image into transformed image.


# a) Watermark Embedding Algorithm

In watermark embedding process, till now researchers are using either DWT or DCT or both transformations. Here, we are using both DWT and DCT to transform the cover image as transformed image and a new and efficient embedding algorithm which embeds DCT coefficients of watermark image into transformed image to get the watermarked image as shown in Fig 1 . It is elucidated in the following way.


# Algorithm

Step 1: Consider any color image as cover image denote it by 'I'. Get R, G, B channels of cover image 'I'.

Step 2: Apply DWT to blue channel 'B' to get the multi-resolution sub-bands LL1, HL1, LH1, and HH1. When compared to red and green channels blue channel is more resistant to changes.

Step 3: Apply DWT again to HL1 sub-band of B channel and choose HL2 sub-band.

Step 4: Divide the HL2 sub-band into blocks of size 8×8.

Step 5: Apply DCT to the blocks obtained in step 4.  Step 7: To embed the watermark, select any four coefficients in the mid frequency band of each DCT block of HL2 sub-band.

Step 8: Four DCT coefficients of watermark are stored in each DCT block of HL2 sub-band.

Step 9: The DC component from the DCT coefficients of watermark is normalized before embedding.

Step 10: Apply IDCT to the blocks of HL2 sub-band.

Step 11: Apply IDWT for 2 levels.

Step 12: Combine R, G, B channels to get watermarked image 'WI'.

It helps to improve the copyright protection of the cover image and robustness of the watermark in the watermarked image. 


# PERFORMANCE EVALUATION METRICS

To measure the quality of the watermarked image peak signal to noise ratio (PSNR) is used. The quality of extracted watermark is measured using Normalized Correlation (NC) between the extracted and the original watermark. 


# MSE

Where n denotes the number of bits used for color representation and MSE refers to the Mean Square Error between original and water marked image and is calculated with the formula.
MSE = ? ? ? (I[i, j] ? I ? [i, j]) 2 ?? ?? =1 ?? ??=1 ??,??,??

# 3MN

Here, M and N are the height and width of image respectively. I[i, j] denotes the (i, j)th pixel value of the original image and I'[i, j] denotes the (i, j)th pixel value of watermarked image.


# b) Normalized Correlation (NC)

It is one of the metrics used to find the quality of extracted watermark image with respect to the original watermark image. It is found by using the following formula.
NC = ? ? w(i, j)w ? (i, j) N j=1 M i=1 ?? ? w(i, j)^2 N j=1 M i=1 ?? ? w ? (i, j)^2 N j=1 M i=1
Here, w(i, j) is the original watermark, w'(i, j) is the extracted watermark.

IV.


# RESULTS AND ANALYSIS

As per this proposed method, a Lena color image of size 512×512 is considered as cover image as shown in Fig. 3 and Fig. 4 is a gray scale image of skeleton with size 32×32 is chosen as watermark. Step 2: Apply DWT to blue channel 'B' to get the multi-resolution sub-bands LL1, HL1, LH1, and HH1.

Step 3: Apply DWT again to HL1 sub-band of B channel and choose HL2 sub-band.

Step 4: Divide the HL2 sub-band into blocks of size 8×8.

Step 5: Apply DCT to the blocks obtained in step 4

Step 6: Obtain four DCT coefficients of watermark from the four selected coefficients of mid frequency band of each DCT of HL2 sub-band.

Step 7: To get the watermark apply IDCT on the set of DCT coefficients obtained from previous step. 


# Table I : PSNR and NC values

Table 2 shows comparison results of the proposed method with the existing transformation methods such as DWT [5], DWT-DCT [9] and Bior [10] based on the NC value between original and extracted watermark when the watermarked image undergoes any attacks.


# Table II : Comparison Results

This robust watermarking method is proposed for increasing the security of data hiding as well as quality compared with the existing algorithms. Authenticity is incorporated by embedding the DCT coefficients of watermark image into the transformed image. Experimental results based on attacks confirm that the proposed algorithm performs better than the DWT, DWT-DCT and Bior based scheme and robustness is achieved by hiding DCT coefficients of watermark image in transformed image by the frequency transformations as DWT and DCT.
6![Consider any gray scale image as watermark, denote it by 'w' and then apply DCT to the watermark 'w'.](image-2.png "Step 6 :")
1![Figure 1 : Watermark embedding process](image-3.png "Figure 1 :")
![explained as follows and it is shown in Fig 2. Algorithm III.](image-4.png "")
2![Figure 2 : Watermark extraction process a) Peak Signal to Noise Ratio It is a metric which is used regularly to find the quality of the watermarked image. It is calculated by considering the following formula PSNR=log (2 n ?1) 2](image-5.png "Figure 2 :")
31![Figure 3 : Original Image Figure 4 : Watermark image](image-6.png "Figure 3 :Step 1 :")
5![Figure 5 : Watermarked image After applying the proposed method the resultant watermarked image is shown as Fig.5.](image-7.png "Figure 5 :")
		
		

			

We are very much thankful to the various reviewers who helped us to bring the paper effectively. The valuable suggestions of Dr. Ch.Rupa could not be underestimated in several situations of developing the paper. We sincerely thank our other research friends for giving their precious cautions. For the invaluable research facilities provided by the magnanimous management for achieving the work, we convey our deepest thanks.

			

			
			
			

				


* 
	
		Spatial domain robust watermarking scheme for color image
		
			ManikMondal
		
		
			DebalinaBarik
		
	
	
		International Journal of Advanced Computer Science
		
			2
			1
			
			Jan 2012
		
	




* 
	
		Strongly robust and highly secured DWT-SVD based color image watermarking: embedding data in all Y, U, V color spaces
		
			LBaisa
		
		
			Gunjal
		
		
			NSuresh
		
		
			Mali
		
	
	
		I.J. Information Technology and Computer Science
		
			3
			
			April 2012
		
	




* 
	
		A blind watermarking algorithm based on DCT for dual color images
		
			ZhengChaomei
		
		
			LiYuan
		
	
	
		International Conference on Computer Science and Network Technology
		
			3
			
			December 2011
		
	




* 
	
		Non-blind watermarking scheme for color images in RGB space using DWT-SVD
		
			NVDharwadkar
		
		
			AmberkerB BGorai
		
		
			A
		
	
	
		International Conference on Communications and Signal Processing
				
			February 2011
			
		
	




* 
	
		Color image watermarking algorithm based on DWT-SVD
		
			LiCheng-Qun Yin
		
		
			An-QiangLi
		
		
			LiLv
		
		
			Qu
		
	
	
		Proceedings of the IEEE International Conference on Automation and Logistics
				the IEEE International Conference on Automation and Logistics
		
			August 2007
			
		
	




* 
	
		A color image watermarking algorithm resistant to Print-Scan
		
			GuoChengqing
		
		
			GuoaiXu
		
		
			XinxinNiu
		
		
			YixianYang
		
		
			YangLi
		
	
	
		IEEE International