Rho-Domain for Low-Complexity Rate Control on MPEG-4 Scalable Video Coding

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.