点击次数：

论文类型：会议论文

发表时间：2011-10-18

收录刊物：EI、Scopus

页面范围：435-441

摘要：Scalable video is an efficient and effective solution to stream video over heterogeneous networks to different clients, where a full resolution video bit stream can be adapted or truncated to meet the diverse network bandwidth requirements. Usually, there are several options to extract video streams for a given network bandwidth. To guarantee the high quality of service (QoS) or quality of experience (QoE) (for example, motion smoothness, etc) at a receiver, we propose an Equivalent Mean Square Error (Eq-MSE) scheme which is developed based on spatial and temporal frequency analysis of input video content. It is known that motion intensive content (like sports video) typically requires higher frame rate for smooth playback, and a relative lower frame rate is sufficient to provide decent visual quality for stationary videos. Proposed Eq-MSE is used to derive such minimal frame rate (MinFR) for different videos to guarantee motion smoothness. A simplified rate model is further introduced to obtain the quantization parameter (QP) given the MinFR, model parameters and network bandwidth. Thus, model derived QP and MinFR are employed to extract the proper video sub-stream from a full resolution scalable stream. It is noted that our proposed model based scalable adaptation is video content dependent. Compared with the default scalable video adaptation without considering the video content impact, our proposed scheme can provide better perceptual video quality by conducting the subjective video quality assessment.