Skip to Main Content
Scalable video coding (H.264 SVC) is an attractive option for video service providers due to its ability to adapt a video's bitstream at the server to suit different network conditions and device characteristics. Lowering a video's bitrate can be achieved through reductions in frame rate, spatial resolution, and/or by increasing the quantization levels applied to the video sequence. In this paper, we first evaluate the effects of such scalability using some full-reference and no-reference video quality metrics, namely PSNR, SSIM, blocking, and blurring. No-reference metrics have the ability to capture the degradation in video quality caused by employing scalability in one or more dimensions. We study if conclusions drawn in previous works, which are based on well-known test video content, hold true for real-world broadcast content. We then discuss how, using these results for a particular content type, the use of no-reference metrics can be enabled in place of, or to supplement, existing widely used full-reference quality assessment metrics. We conduct an experimental analysis by transmitting video encoded at different scalability points over a lossy network to ascertain the effect of loss when scalability is employed in one or more dimensions. We analyze these results using a reduced reference metric called delta-blocking, which can detect visual damage of frames that causes a decrease in a user's quality of experience when perceived by the user. If the levels of packet loss are excessively high, this can lead the decoder to drop some video frames. To combat this type of frame loss, we propose a simple windowing algorithm that can automatically re-align the corresponding values for reduced-reference comparison, allowing for video quality monitoring to continue.