CiiT International Journal of Digital Image Processing, Vol 3, No 1, January 2011 34 0974-9691/CIIT–IJ-1213/09/$20/$100 © 2011 CiiT Published by the Coimbatore Institute of Information Technology Abstract---Digital watermarking is intended to protect digital media against unauthorized usages. The watermark is embedded in streamed videos for copyright protection purposes. In video streaming, the watermark embedding and detection time is the main concern. The distribution of the watermarking key is another important issue. The research aims to present a real time watermarking scheme for copyright protection in H.264 compressed videos. The proposed scheme is invisible, blind, robust, and preserves the original video bit rate. The watermarking process doesn’t require any key distribution. In this paper, a new H.264 model is introduced. The context adaptive variable length codes (CAVLC) blocks are statistically modeled. Based on the resulted CAVLC model, some of the CAVLC blocks are selected for watermarking using a user defined threshold. The watermark is embedded in the CAVLC block structure. In watermark embedding, the block’s total coefficients number (TCN) is modified. The simulation results show that, the average embedding and extraction time is zero. Moreover, high bit capacity is achieved compared to other work. The PSNR values of the watermarked videos are up to 40 dB. The embedded watermark is invisible, robust, and the embedding locations are totally hidden from the attacker. The detection algorithm is blind. Furthermore, the total file size is decreased by 8% after watermarking. So, the algorithm is more appropriate for internet applications. Keywords---Copyright Protection, Real Time Video Watermarking, H.264 Watermarking, CAVLC Modeling, VLC Watermarking, Watermark Embedding, Watermark Extraction. I. INTRODUCTION .264 represents an evolution of the existing video coding standards. It is developed to meet the growing needs for higher compression of moving pictures. The standard is jointly developed by the ITU-T VCEG and ISO/IEC MPEG [1], [2]. It achieved higher bit rate and lower distortion ratio compared to MPEG-2. Moreover, H.264 supplies new features in prediction modes. DCT transforms, quantization and entropy coding are also enhanced in the standard [3]. In this paper, the standard is referred to as H.264. H.264 provides two entropy coding modes. The first mode is the context-based adaptive binary arithmetic coding (CABAC). It is used to code all syntax elements. The second mode consists of two entropy coding tools. The first tool is the context-based adaptive variable length coding (CAVLC) which is only used to Manuscript received on December 11, 2010, review completed on December 20, 2010 and revised on December 24, 2010. T. M. Mohamed is with the Department of Information Technology, Faculty of Computers and Information, Cairo University,#5 Dr. Ahmed Zewail Street, E-Mail: tahamahdy3000@yahoo.com H. N. Elmahdy is with the Department of Information Technology, Faculty of Computers and Information, Cairo University,#5 Dr. Ahmed Zewail Street, E-Mail: ehesham@fci-cu.edu.eg H. M. Onsi is with the Department of Information Technology, Faculty of Computers and Information, Cairo University,#5 Dr. Ahmed Zewail Street, E-Mail: h.onsi@fci-cu.edu.eg Digital Object Identifier No: DIP012011006. encode quantized coefficients. The second tool is the universal variable length coding (UVLC) which is adopted to encode other syntax elements. Generally, CABAC watermarking is not suitable for real time requirements because it requires more calculations than CAVLC. Alternatively, CAVLC watermarking is more appropriate [4], [5]. The watermark is extra information embedded into digital data. Transparency, robustness, capacity are the most important properties required for watermarking systems. Moreover, video streaming applications require real time watermarking. The surviving of watermarks against lossy compression, such as H.264, is an additional feature [6], [7]. Most of the watermarking requirements are conflicting with each other. The design of the watermarking system should optimize these requirements depending on the intended application [8]. In the compressed domain, the watermark embedding process may be performed in DCT or quantized DCT or VLC domains [7]. DCT watermarking is a direct modification of selected DCT coefficients. The modification is often performed using spread spectrum concepts. It is the most robust domain against watermarking attacks. However, the watermarking process is not suitable for real time applications. The quantized DCT watermarking is the process of modifying selected quantized DCT coefficients for watermark embedding. In case of watermarking, this domain is considered as a moderate with respect to real time and robustness. VLC watermarking is performed by modifying VLC codes in the compressed video. It is preferred in real time watermarking applications because only VLC encoding and decoding is required for watermark detection [7]. However, the watermarking robustness in this domain is limited. This is due to that, the actual coefficients values are ignored during watermarking process. So, any bit change in a VLC code, causes a decoding ambiguity. This ambiguity is occurred if the watermark is embedded or not. Another disadvantage of VLC watermarking is the possibility of exchanging one VLC with another. The process causes watermark destruction. However, H.264 provides a set of error resilience tools that are used for video recovery. The recovery process is an inclusive recovery of the watermark [6], [9]. Generally, the VLC watermarking in MPEG-2 and all of the previous standards is a mapping process of one VLC to another. So, the embedding process is straightforward. Unfortunately, MPEG-2 watermarking algorithms can’t be directly applied to H.264. A glance on H.264 shows that, the components of a CAVLC block are highly related. So, altering any CAVLC component may cause inconsistencies in the block structure. These inconsistencies lead to a block decoding ambiguity [9]. To illustrate this issue, a CAVLC block is composed of seven components {Total Coefficients Number (TCN), T1, T1 Sign, Real Time Watermarking of H.264 Video Sequences Taha M. Mohamed, Hesham N. Elmahdy and Hoda M. Onsi H