A Video Reconstruction Approach Supporting Frame Skipping in H.264

Mobile real-time video streaming for web-casting and video conferencing conforming to the ITU-T H.264 data compression recommendations format is emergent. However, the quality of the video services for typical applications is hard to foreseen and sustain. Qualities of Services (QoS) approaches, such as video frame skipping techniques, are thus attractive to address the problem. Many conventional video frame-skipping techniques focus on the system-centric quality attributes to adjust the quality of video delivery. They normally ignore the user-centric quality attributes. A usercentric approach, amongst others, is to take human perception in consideration and adapt the video quality expressed in patterns of Group of Pictures (GoP). In a post-processing step of an encoding stage, it filters and re-titles the qualitymatching patterns from the rest. It then reconstructs the filtered GoP according to the H.264 format. Most existing H.264 playback engines unfortunately cease to work properly for video steaming with skipped frames. This paper reports our experience on alleviating the playback problem through a case study on our effort to develop a prototype of such a playback engine. It studies how to reconstruct a video steaming in the H.264 format on the fly in the presence of predictable skipped frames, satisfying certain selected user-centric QoS quality levels. We also analyze the applicability of the approach.