Digital Color Image Watermarking Using Encoded Frequent Mark

With the increased development in digital media and communication, the need for methods to protection and security became very important factor, where the exchange and transmit date over communication channel led to make effort to protect these data from unauthentication access. 
This paper present a new method to protect color image from unauthentication access using watermarking. The watermarking algorithm hide the encoded mark image in frequency domain using Discrete Cosine Transform. The main principle of the algorithm is encode frequent mark in cover color image. The watermark image bits are spread by repeat the mark and arrange in encoded method that provide algorithm more robustness and security. The proposed algorithm efficiency is measured by using many of measurement factors such as Peak Signal to Noise Ratio PSNR and Normalized Correlation Coefficient NC, the watermark robustness and feasibility are measured by using many types of attacks. 
 

The main objective in the proposed paper is to present the hiding watermark in color cover image using encoded frequent mark embedded in frequency hopping since much than one frequency are used to embed the watermark in DCT blocks for get more security , then the watermark image is extracted at the receiver without need for origin image .The rest of this paper is organized as follows: In Section 2 Discrete cosine Transform explained.Section 3 discusses the Basic watermarking criterias.Watermark Embedding and Watermark Extraction respectively explained in section 4 and section 5. Error and Correction Ratio are explained in Section 6. Section 7 contain Encoded and Frequent Stage.Simulation Results are discussed in Section 8. Finally, Section 9 concludes the paper.

BASICS WATERMARKING CRITERIAS
Watermarking algorithm to be efficient, many criterias should be obtained or saved as explained:-1-Invisibility: which means the quality of cover image should not be effected and cannot be explained by users Giri, et al., 2015.3-For the mark, encoded frequent mark is used, where the binary image mark is used and the mark is encoded in way that depend on repeated the bits value of mark and reposition these bit in a sequence predefined at Extraction side.This step consider the main aim for this algorithm, where the criterias of watermarking image are saved.The number of repeated is commensurate extrusive with capacity criteria, where as directly proportional with security and robustness criterias.

2-
4-The encode bits are insert in frequency domain layer of cover image, middle frequency positions are used for insertion and these positions selected using predefined secret key.

5-Convert the Luminance component (Y) from Frequency domain to spatial domain.
Convert YCbCr components to RGB components to obtained watermarked image.

PROPOSED WATERMARK EXTRACTION
At the receiver side in Fig. 2 backward steps are used to recover mark image as explain: 1-Convert the watermarked color image from RGB components to YCbCr components.

2-Convert the Luminance component (Y) to Frequency domain to.
3-Select the position of insertion mark value, where the middle frequencies are used and these positions can be selected depend on secret key.
4-After extract the mark value, decode stage is used which include obtaining a one copy of mark image.The number of copy and rules of select sequence mark value are predefined in Embedding stage.5-The value of the sequence is threshold to '1' or '0'.

ERROR AND CORRECTION RATIO
The quality of the watermarked image can be calculated by finding the value of Peak Signal to Noise Ratio PSNR Sharma, and Prashar, 2012, as shown: Where the RMSE is root mean square error and equal to: (2) M and N are image dimensions (number of pixels) (M=N).The similarity between extracted mark and the origin mark can be calculated by finding the Normalized correlation coefficient NC as in Eq. ( 3): (3) Where w (i , j) and w*(i , j) represent the origin and extracted watermark respectively.The value of NC equal to1 as maximum value Zubair , et al. ,2009.

ENCODED AND FREQUENT STAGE
This stage include two steps, the first step make numbers of copy of mark and this number is commensurate extrusive with security and robustness criteria where as inversely proportional with the Capacity and PSNR for the watermarked image.Fig. 3 show the ratio between the PSNR and the number of mark copies.Second step include redistribution the bits sequence of the copies in away predefined at embedding and extraction stage.

SIMULATION RESULTS
In order to test the proposed algorithm, color (512*512) pixels images are used.Lena, Peppers and Airplane image used as cover images.In other side binary images size (64x64) pixels is used as mark image.The size of (512*512) constribte in capacity criteria.PSNR calculated to depend it as criteria metrics (it calculated for Number of mark copy =2).Table 1 show the value of PSNR While Table 2 show Normal Correlation and how it effect during transmitted or attack by hackers and these process may changed or modified.

CONCLUSION
New proposed algorithm presented in this paper for watermarking digital color image.This algorithm focus on away where make the criterias of watermarking are most important properties.Color image cover has three layers of color and one of these layer is used.The number of copies that can be inserted in this layer may be one copy or more ,this depend on the value of PSNR for watermarked image where the PSNR will reduce when the number of copies increased as shown in the graphs in Fig. 3 and the property of capacity achived .The redistribution of these copies give a security form for algorithm since the position of bits injection are known only in the extraction stage and it will contibute in increasing the security criteria .Techniques are used such as the position of insertion in DCT domain and position of insertion in over all the image led to reduce the effect of attacks on the watermarked image and as a result the robutness became agood.These properties are explained in results obtain as shown in Table 1 and 2 The process of embedding invisible signal in an image called Digital Watermarking Image.This process consider a form of copyright Protection, ElGamal, et al., 2013.Watermarking classify according to many parameters, at first according to the domain of embedding and extraction where two domain are used spatial domain and frequency domain.The spatial domain used the cover image without convert it to frequency domain and it is characterized by simplicity and less immunity to attacks where as in frequency domain the cover image transfer to frequency domain using Discrete Cosine Transform (DCT) or Discrete Wavelet Transform (DWT), the transform coefficients modified by the mark value, then to get the mar image, one should perform the transform inversely, Hajjara , et al. , 2009.The visibility is another parameter to classification where there are two types: visible watermarking and invisible watermarking, visible water mark is clearly detectable .While the invisible watermarking is used to identify the owner or the origin of the host image, Giri ,et al. ,2015 .Many techniques for color image watermarking are presented, Some of these techniques focus on spatial domain as in Dharwadkar, and Amberker, 1999, where the paper focus on the average intensity which is estimated from the intensities of the R, G and B components.In Abdallah, and Hammo , 2006, deals with colored images in the biological color model, the Hue, Saturation, and Intensity (HSI), where the method robustness is tested against several attacks.Watermark security is increased by using the Hadamard transform matrix ..Another part of researches focus on the transform domain using (DCT) as in Abraham, and Paul, 2016, it uses Discrete Cosine Transform to transform the host image into frequency domain for selecting appropriate coefficients for watermark embedding.Wavelet transform is another type of the transform domain which provide the property of multi-Resolution analysis, many of searcher focus on this property as in Sharma ,and Prashar, 2012 .There are papers merge more of transform as in ElGamal , et al. , 2013, the searchers joined between DCT and (DWT) which take advantage of the two frequency domains.
Discrete Cosine Transform (DCT) can be defined is a transformation algorithm that converts images from spatial domain to frequency domain ElGamal , et al. ,2013.Carried out watermarking need to change the transform coefficients and the DCT watermarking is more robust compared to the spatial domain Hajjara, et al. , 2009 .
Robustness: Means the immunity to the various signal operation that can be obtain during transmit such as noise ,compression and resizing Patel ,and Tahilramani ,2016.3-Security: can be defined as the ability to resist hacking attempts to remove the mark without modified cover image Hui-fang, et al., 2010.4-Capacity: means amount of information bits can be inserted in cover image ElGamal, et al. , 2013. 4. PROPOSED WATERMARK EMBEDDING As shown in Fig.1, many stages are used in embedding side, these stages explained as: 1-Convert the color image from RGB components to YCbCr components.Since the Luminance (Y) is less sensitive for human.2-The Luminance (Y) layer convert to (8*8) non-overlapped DCT domain.The size of (8*8) standard as in Abraham, and Paul, 2016.

Table 1 .
PSNR for Watermarked Color Image.