H . 264 / AVC and SVC Video Trace Library : ∗

There are three ways to characterize encoded video for the purpose of networking research: (i) video traffic model, (ii) video bit stream, and (iii) video traffic trace. Video traffic models strive to capture the essential properties of the real traffic in parsimonious mathematical models. Video traffic models are typically developed based on the statistical properties of samples of the real traffic, or, in many cases, video traces of the real traffic. Video traces are therefore typically a prerequisite for model development. Video traffic models for H.264/AVC and H.264 SVC are largely an open research area. The bit stream contains the complete video information and allows for networking experiments where the quality of the video—after suffering losses in the network—is evaluated (a capability that can be emulated with offset distortion video traces [6]. Key limitations of the bit streams are that they require expertise in video coding, are very large in size, and are usually proprietary and/or protected by copyright. Video traces are an attractive alternative to traffic models and bit streams in that they represent the traffic and quality of the videos. While the bit streams contain the actual bits carrying the video information, the traces are simple files that contain the number of bits needed to encode each video frame and the quality level (e.g., in PSNR) of the encoding [7]. Thus, there are no copyright issues with traces and networking researchers do not need to acquire detailed video coding expertise nor video coding equipment to conduct video networking research.