Content based video quality estimation for H.264/AVC video streaming

The scope of this work is the estimation of video quality for low resolution video sequences typical in (mobile) video streaming applications. Since the video quality experienced by users depends considerably on the the spatial (edges, colors,...) and temporal (movement speed, direction,...) features of the video sequence, this paper presents a two-step approach to quality estimation. Firstly, shots between two scene changes are analyzed and their content class is found. Secondly, based on the content class, frame rate and bitrate, an estimation of quality is carried out. In this paper, the design of the content classifier as well as an appropriate choice of the content classes and their characteristics is discussed. Moreover, the design of quality metric is presented, based on the mean opinion score obtained by a survey. The performance of the proposed method is evaluated and compared to several common methods. The results show that the proposed approach, provides powerful means of estimating the video quality experienced by users for low resolution video streaming services.