A DISCRETE COSINE TRANSFORM BASED WATERMARKING SCHEME FOR COLOR IMAGE USING YCbCr SPACE

In this paper, a non-blind color image watermarking scheme is proposed based on discrete cosine transform and YCbCr color space. The host image is converted from RGB space into YCbCr space and the Y component image is employed for embedding a binary watermark image. The strength of the watermark is controlled by a robustness factor to obtain an acceptable trade-off between robustness and imperceptibility. The performance of the proposed scheme is evaluated in term of Peak signal to noise ratio (PSNR) and normalized correlation (NC). The experimental results show that the proposed scheme is robust under various attacks like JPEG compression, filtering and noise addition.


Introduction
With the quick advancement of computer and network technology, computerized information can now be transmitted through internet quick and simple.Since the digitized information could be unlawfully copied and effortlessly altered, the enforcement of multimedia copyright protection became an essential issue.One of the effective approach to take care of this issue is digital watermarking for copyright protection.Digital watermarking is the process of embedding information (watermark) into a multimedia content (audio, image or video) in some way the embedded information can be identified or extracted later for the purpose of copy right protection, authentication, ownership verification, broadcast monitoring, etc.The quality of watermarking scheme is determined by four major factors including robustness, imperceptibility, security and capacity [1].A good watermarking scheme should be robust against a variety of possible attacks like scaling, compression, cropping, adding noise, filtering and others without distortion the host image quality (imperceptibility).Thus, there is a need to improve watermarking algorithms that can offer good balance between imperceptibility and robustness.Generally, image watermarking methods can be implemented either in spatial domain or in frequency domain.In spatial domain, the inserting of watermark is done by directly modifying the pixels value in such a way that remaining the watermark is invisible.Least significant bit (LSB) is one of the most commonly used algorithms in spatial domain.In transform domain, the pixel value is substituted with the transform coefficient and the embedding process is done by altering the transform domain coefficients.Most commonly used transform domain algorithms are discrete cosine transform (DCT) and discrete wavelet transform (DWT) [2].The transform domain techniques are more effective than the spatial domain for achieving the robustness and imperceptibility, as clarified by various surveys [3][4][5].Hence, the transform domain is utilized for the proposed scheme.

Discrete Cosine Transform
The Discrete Cosine Transform (DCT) is an orthogonal transformation that is very widely used linear transform in digital signal processing.DCT decomposes a signal into a series of symmetric cosine functions therefore the obtained matrix is determined by the horizontal, vertical, and diagonal frequencies.The key feature of the DCT decomposition is that the vast majority of the signal information tends to be collected in a few low-frequency components of the DCT and the part that the eye of human is least sensitive to is neglected.Thus, it is very widely used in image compression and is widely accepted in the multimedia standards.The two dimensional discrete cosine transform and inverse discrete cosine transform (IDCT) for an image is described by the following equations [6]: (3)

YCbCr Color Space
The digital image watermarking is often accomplished in gray scale but color image watermarking can be employed to improve the performance of the watermarking scheme.Color spaces are three-dimensional systems used to represent specific organization of colors depending on three coordinates that describe color.The RGB space is the basic one, that can be transformed into other color spaces.YCbCr color space represents each color with three components.The Y component describes the light intensity (luminance).The Cb and Cr components are describe the color information.Thus, YCbCr color space separates brightness and chroma.In the proposed scheme, the Y component is used for watermark embedding process to satisfy robustness quality measures.The RGB to YCbCr conversion is defined as [7]: The inverse conversion from YCbCr to RGB space is defined as:

Proposed Scheme
In the proposed scheme, a binary watermark image is embedded in color image by using DCT and YCbCr color space.The proposed scheme employs the properties of DCT and YCbCr space to increase the robustness and imperceptibility of the watermarking algorithm by using the most significant coefficients of the whole Y component image.The watermark embedding and extraction algorithms are clarified in the following sections and illustrated in Figs. 1 and 2 respectively.

Watermark Embedding Algorithm
The watermark embedding algorithm is described by the following steps: Step 1: Read the original color image I with the size of and the binary watermark image with the size of .
Step 2: Convert the RGB space of the original image into YCbCr space and select the Y component.
Step 3: The DCT for the (Y component) is calculated to produce matrix.
Step 5: Take an element ( ) of the watermark image, embedded in the selected coefficients of according to the following formula: Where is the robustness factor, the value of is discussed in section 5.
, where: Step 6: Apply the IDCT to the altered Y component and combine the YCbCr space components Step 7: Obtain the watermarked image by switching the color space from YCbCr to RGB.

Watermark Extraction Algorithm
The watermark extraction algorithm is described by the following steps: Step 1: Read the original image and watermarked image .
Step 2: Convert the RGB space of the original and watermarked images into YCbCr space and select the Y components .
Step 3: Apply the DCT to the whole Y components to obtain y and y'.
Step 4: Obtain the (n) maximum DCT coefficients T ' = {t' 1 , t' 2 ,…, t' n } of y' according to the DCT coefficients T = {t 1 ,t 2 ,…,t n }, where the DCT coefficients of T and T ' have the same positions in the DCT matrices of and respectively.

Experimental Results and Simulation
The quality of the digital watermarking techniques is commonly estimated by the robustness of the watermark to the common signal processing operations of the watermarked image and the imperceptibility of the embedded watermark to human observers.In this paper, the proposed scheme is evaluated for both properties of robustness and imperceptibility by using Matlab platform.Three 512 512 well-known color images: Airplane, Peppers, and Baboon shown in Fig. 3(a-c) are used as the host images for embedding a 30 40 binary watermark image, shown in Fig. 3(d).scheme, the Normalized Correlation (NC) is used to measure the similarity degree between the original watermark image ( ) and the extracted watermark .When the NC value is 1, means that the original and the extracted watermark is precisely similar.NC formula for two images is given by [8]: For evaluating the imperceptibility of the embedded watermark, the peak signal to noise ratio (PSNR) is used to measure the quality of the watermarked image.
The higher the PSNR value the lower the perception of the watermark in the host image.PSNR is given by [8]: Where the MSE is the mean square error and defined as: For achieving an acceptable tradeoff between robustness and imperceptibility, the strength of the proposed watermarking scheme is controlled by a factor ( ).In the other word, the value of determine the strength of the watermark on the host image.theoretically, for high values of , the robustness of the watermarked image increases while its quality decreases and vice versa.Fig. 4 shows this effect visually according to human observer and Tables 1 and 2 show this effect in quantitative manner.Table 1.Performance comparison1 with varying values of α (from 0.01 to 0.2).
Table 2. Performance comparison2 with varying values of (from 0.01 to 0.2).
Tables 1, 2 and Fig. 4 show that the value of ( ) should be chosen carefully to obtain a good balance between the robustness of extracted watermark and the imperceptibility of watermarked image.In these tables, the PSNR is calculated by obtaining the average PSNR value for the red, green and blue channels [9].Fig. 5 and 6 show the watermarked peppers images and the detected watermark logo images after subjecting the watermarked image to different known attacks with the corresponding values of PSNR and NC.

Conclusions
This paper offers a non-blind watermarking method for color images based on DCT and color space conversion.Since the YCbCr color space separates brightness and chroma, the use of Y component for watermark embedding is led to improvement in results.The watermarked image is exposed to different attacks (JPEG compression, Gaussian noise, salt and peppers noise, speckle noise, median filtering, average filtering, cropping and histogram stretching) for performance evaluating.Figs.4(b) and 4(c) show that the proposed scheme attains the imperceptibility of the embedded watermark against the human observer .According to Tables 3-5, it is obvious that the proposed scheme achieves both the robustness and the imperceptibility under different attacks.

Figure 4 .
Figure 4.The effect of value on the watermarked image quality.

Figure 5 :Figure 6 :
Figure 5: Watermarked Peppers images subjected to different image processing operations attacks and the extracted watermark logo image of each one with 0 .

Table 3 .
Performance evaluation and comparison for Airplane image.