Scalable video coding was designed in response to the growing need for flexibility in video transmission over networks and channels. MPEG-4 scalable video coding (SVC) is a recently finalized standard which introduces new coding tools such as spatial, temporal and quality scalability, to produce a layer-based scalable video stream. Additionally, inter-layer prediction allows a layer to use information from other layers as a basis for motion and texture prediction, improving the overall coding efficiency. Rate control is a capital issue in video coding, as it is designed to regulate the bitrate at the output of the encoder and keep it close to a specified constraint. Whereas rate control has been extensively studied for non-scalable video coding, only few propositions were made for scalable video coding. In this paper, we adapt an attractive rate control approach, based on a bitrate modeling framework called rho-domain, for scalable video coding. We show that this model performs well on all spatial, temporal and quality scalabilities, and handles inter-layer prediction quite accurately. After validating the approach in MPEG-4 SVC, we use the rho-domain model to build a simple-accurate rate control scheme. Results show that the mean frame bitrate error is below 7% on a representative set of configurations, while the impact on the complexity of the encoder is very low.
Published in:
Multimedia, 2008. ISM 2008. Tenth IEEE International Symposium on
Date of Conference: 15-17 Dec. 2008