The objective of this paper presentation to explain how to embed the spatial spread-spectrum watermark directly to compressed MPEG-4 bit streams by modifying DWT coefficients. The proposed watermarking procedure is based on the three-dimensional discrete wavelet transform (3D DWT) and the spread spectrum sequence. First the watermark image has been preprocessed using the mixing and the pseudorandom permutation. After dividing a video sequence into video shots, perform the 3D DWT. And then, the watermark is proposed to embed into the 3D DWT coefficients considering the robustness and the invisibility. At this time, two spread spectrum sequences define as the temporal user key has been used. The watermarked frames should be indistinguishable from the original frames subjectively. The proposed video watermarking algorithm has to test under different attacks such as the low pass filtering, the frame dropping, and the MPEG coding
Keywords: MPEG, Discrete wavelet transform, watermark, Quantization,
1 INTRODUCTION
Digital watermarking is the process by which, identifying data is woven into media content such as images, movies, music or programming. Imperceptible to the human senses, yet easily recognized by special software detectors, a digital watermark gives content a unique identity that remains constant even through recording, manipulation and editing, compression and decompression, encryption, decryption and broadcast without affecting the quality of the content. Supporting a number of different business models and environments, digital watermarks are easy to apply and difficult to detect without the appropriate technology. As a result, they present a roadblock for those who attempt to pirate digital content. On the other hand, these same characteristics make it easier and less expensive for content owners to place unique identifiers within their digital content as a means to manage and track that content as it’s distributed. Since the digital watermark remains with the content even as formats change it presents new opportunities for content owners to explore new digital business models without sacrificing their rights.
Digital watermarks can be used in video, audio, print and digital images for a wide variety of applications such as copy prevention and play control, monitoring and tracking; copyright communication, rights management, authentication and enabling consumer access to and filtering of Internet content. The objectives of such watermarks are completely different: A (semi-)fragile watermark is a mark which is (highly) sensitive to a modification of the stego-medium. A fragile watermarking scheme should be able to detect any change in the signal and identify where it has taken place and possibly what the signal was prior to modification.
2 METHODOLOGY
2.1 THE VIDEO COMPRESSION CODING BASED ON 3-D DWT
As shown in Figure, video coder consists of three major modules wavelet transform, motion compensation, and quantization. Wavelet decomposition can be operated either on the original video samples before the motion compensation or on the residual video samples after motion compensation. The following two are popular schemes that have been implemented:
2.2 WATERMARK EMBEDDING
In the proposed video watermarking method, first, to divide the video sequence into video shots, spatial different metric (SDM) has been used. This is to use the dissimilarity between the adjoining frame pair, this method is that the efficiency is low but the algorithm is simple. And then, the spatial 2D DWT is performed and the temporal 1D DWT is performed about selected video shots respectively. In the 3D DWT coefficients, to embed the preprocessed watermark images into the HL subband and the LH subband of three level about the spatial axis and lowpass frames about the temporal axis. LL subband of three levels is to satisfy the invisibility and highpass frames consisted of the dynamic components to satisfy the robustness.
The watermark ) is embedded using the following relationship
}
Where Vf (i, j) denotes the 3D DWT coefficient of the watermarked frame. Vf(i, j) denotes the 3D DWT coefficient of the original frame. α is the scaling parameter. The watermark is not the binary watermark but one of two independent two spread spectrum sequences. Two independent sequences is defined as the temporal user key and also can be the secrete key with the visible binary watermark.
Watermark is embedded to video frames by changing position of some DWT coefficient with the following condition:
If W[j] = 1,
Exchange C[i] with max(C[i], C[i+1], C[i+2], C[i+3], C[i+4])
Else
Exchange C[i] with min(C[i], C[i+1], C[i+2], C[i+3], C[i+4])
Where Ci is the ith DWT coefficient of a video frame, and Wj is the jth pixel of a corresponding watermark image. When the watermark Wj =1, we perform an exchange of the Ci with the maximum value among Ci, Ci+1, Ci+2, Ci+3, Ci+4. When Wj = 0, we perform an exchange of the Ci with the minimum value among Ci, Ci+1, Ci+2, Ci+3, Ci+4. With this algorithm, the retrieval of the embedded watermark dose not needs the original image. The sequence of watermark coefficients used is depicted in Figure 2.1, where higher frequency coefficients are embedded to higher frequency parts of the video frame and only the middle frequency wavelet coefficient of the frame (middle frequency sub band) is watermarked.
Fig.2.1.Embedding watermarks in a frame
Video embedding algorithm:
(i) The host video is segmented into scene shots, and then some shots are selected by randomly for embedding watermark. For each selected scene shots, the following steps are repeated;
(ii) Perform a 2D wavelet on each frame Rk, (k = 1,2,…., n) , the wavelet coefficient frames are denoted by Xk,(k =1,2,...,n) ; select the watermark embedding regions in each Xk (k = 1.2,. . . , n) , then perform a 1D WT across the temporal axis (between frames), and E, denote the 3D wavelet coefficient frames;
(iii) Watermark spread spectrum embedding is based on magnitudes of the 3D DWT coefficients. Watermark image is adaptively spread spectrum and embedded into these coefficients. For each pixel (i,j) of the selected area in Rk(k = 1,2,. . . , n) , the value is compared with those of its eight neighbors; t denotes the total number
which the value is larger than its neighbors. Watermark is embedded by changing the corresponding coefficient value
(iv) By the inverse 3D DWT, the watermarked video is obtained. Watermark image is embedded into video by adaptive spread spectrum based on the characteristic of video data, in which a watermark image W is changed into an image sequence Wk(k =1,2,...,n) . So each pixel in the original watermark image is embedded into every compositions of 3D WT Coefficient in different degree, including low-pass frames and high-pass frames. It is to say that each watermarking data is related to the static and the dynamic composition in the shot.
2.3 WATERMARK DETECTION
The watermark extracting process is the inverse procedure of the watermark embedding process and the similarity to extract the final watermark image has been used. The proposed algorithm requires the original video sequence and the user key. The difference between the wavelet coefficients of the original frame and the watermarked frame has been obtained. The difference is not the extracted binary watermark images but the extracted temporal user key modified by attacks.
As an identical watermark is used for all frames within a scene, multiple copies of each part of the watermark may be obtained. The watermark is recovered by averaging the watermarks extracted from different frames. This reduces the effect if the attack is carried out at some designated frames. Then combine the 8 bit planes and recover the original size image, i.e., half part of the original watermark Accordingly, the similarity between the original temporal user key and the extracted temporal user key has been calculated.
This is to extract the binary watermark image. The extracted watermark image W' (i,j) is
Where xf (i, j) is the extracted temporal user key value.
The similarity is
As the 2D reverse pseudorandom permutation and the reverse mixing was performed, the visible binary watermark image has been extracted. The watermark is a binary image. Therefore, the viewer can compare the extracted watermark with the original watermark subjectively. However, as the subjective measurement depends on the condition of the viewer. The normalized correlation (NC) as the objective measurement has been used as subjective metric. The NC is
Where w (i, j) is the original watermark image and w'(i, j) is the extracted watermark
Image
.
3 RESULTS
Computer simulations have been carried out to demonstrate the performance of the proposed algorithm. The standard sequence flower garden video sequence has been used as the test video sequence. The sequence , in CIF format (frame size: 352 x 288 pixels, progressive scan, 4:2:0 sub sampling ( format ) is used in the experiment. The MxNxT DWT cube contains information regarding each of the T image frames.
When some intentional or unintentional attacks have been applied to the watermarked video, the corresponding experiment results are shown as follow section
Fig. Performance of proposed algorithm Fig. Performance of proposed algorithm
under noise attack under averaging attack
Fig. Performance of proposed algorithm Fig. Performance of proposed algorithm
under frame averaging attack under frame average attack
4 CONCLUSION
The process of this video watermarking scheme, including watermark preprocessing, video preprocessing, watermark embedding, and watermark detection, is described in detail. The robustness of the proposed video watermarking procedure has been illustrated to several video degradations, including noise attack, video frame dropping and video frame averaging.
5 SCOPE FOR FUTURE STUDY
For the video watermark detection, the raw video is needed which is not convenience for many cases. So, the blind watermarking detection algorithm should be investigated in the further research. Moreover, we need to further optimize the video watermark embedding and detection procedures so that it can be performed in real time. The scheme can be improved by making use of the information from the video, such as time information, to increase the robustness of the watermark
6 REFERENCES
[1] Seung-Jin Kim , Suk-Hwan Lee, Kwang Seok Moon, Woo Cho “ A new digital
video watermarking using the dual watermark images and 3D DWT ” IEEE Trans.
on Image Processing , Vol.44 , pp. 708-712, Aug. 2004.
[2] Hongmei liee, Nuoe Chen,Hwang “ A robust 3D wavelet video watermarking IEEE Trans. on Image Processing , Vol. 32, pp. 208–212, Oct. 2003..
[3] Ying Le,Xinbo Gao,Hongbing Ji “A 3D wavelet based spatial-temporial approach
for video watermarking”, Proc. IEEE Computational Intelligence and Multi media
applications Conf., pp. 244–250, Mar 2003.
[4] X. B. Gao, X. 0. Tang “Unsupervised video - shot segmentation and model-free in anchorperson detection for news video story parsing”IEEE Trans. On Circuits and Systems for Video Technology, Vol.4,pp 1654-1658,Dec 2002.
[5] X. Niu, S. Sun, and W. Xiang “Multi resolution Watermarking scheme for Video based On Gray- Level Digital Watermark,” IEEE Trans. Consumer Electronics, vol. 46, pp. 375-384, May 2000.
[6] C. T. Hsu and J. L. Wu, “ Hidden Digital Watermarks in Images,” IEEE Trans. Image Processing, vol. 8, pp. 58-68, Jan. 1999.
[7] M. D. Swanson, B. Zhu, and A. H. Tewfik, “ Multiresolution Scene - Based Video
Watermarking using Perceptual Model analysis,” IEEE Journal Selected Areas in communication, vol. 16, no. 4, pp. 540-550, May 1998.
[
ReplyDeletevery good article very nice article I inspired by this article.
Next page