Pengpeng Ni

Visual Perception of Scalable Video Streaming: Applied to Mobile Scenarios

Supervisor(s) and Committee member(s): Pål Halvorsen (supervisor), Carsten Griwodz (supervisor), Klara Nahrstedt (opponent), Hendrik Knoche (opponent), Professor Frank Eliassen (opponent)


img_8714_0Modern video coding techniques provide multidimensional adaptation options for adaptive video streaming over networks. For instance, a video server can adjust the frame-rate, frame-size or signal-to-noise ratio (SNR) of the video being requested to cope with the available bandwidth. However, these adaptation operations give rise to distinct visual artefacts, so it follows that they are not perceived in the same manner. Subjective evaluation shows that we can no longer assume a monotonic rate-distortion function for scalable video. In fact, the perceived video quality that is expressed as the overall viewing experience of the video being delivered, is mutually and interactively affected by many factors ranging from configuration parameters to source material.

Performing subjective evaluation is a tedious task and field studies are especially expensive and difficult to administer. This work presents a subjective quality evaluation method for conducting audiovisual quality assessment studies in the field. Our method is designed with realistic assumptions of the time and effort that an assessor will have to spend. With the use of this method, an experimenter can easily obtain stable results with accuracy close to traditional experiment designs at a much lower cost. We demonstrate the efficiency and practicality of this method by simulations.

To help formulating optimal adaptation strategies for streaming services, a sequence of field studies have been conducted to evaluate the perceived video quality, with the focus mainly on mobile streaming scenarios. These studies reveal that frequent quality variations may create additional visual artefacts denoted flicker effects, and it is not worthwhile making quality changes unless the negative impact of flicker on visual quality is eliminated. We identify the main influential factors on the visibility of flicker effects and determine the threshold quantities of these factors for acceptable visual quality of video. These findings can help improving video adaptation strategy or bit-rate controllers deployed in video streaming solutions, such as Scalable video streaming and Dynamic Adaptive Streaming over HTTP. In all of our studies, the quality assessments were made on different types of video content, we therefore provide some preliminary analyses of content effects on human quality perception. In addition, we found that human perception of visual artefacts varies in relation to the viewing environment. Especially, people can detect slow or irregular frame-rates much easier on large HDTV screens than small screens of mobile devices.

Media Performance Group


Bookmark the permalink.