Skip to Main content Skip to Navigation

Stratégies d'encodage pour codeur vidéo scalable

Abstract : This work is about designing bit rate control technics for the MPEG-4 Scalable Video Coding standard. The proposed approach benefits from a low computational complexity, so that its impact on the encoding time is very small. The quality of the encoded stream is also considered when controlling the bitrate in order to obtain optimal visual impression. With today's heterogeneous communication technologies and video-reading devices, broadcasting video content requires large amounts of time and money to provide the user with optimal quality in any context. Scalable video coding has been developed as an answer to this need for adaptive video streams. In 2008, the MPEG-4 Scalable Video Coding (SVC) extension was finalized. To address the adaptivity issues, this new standard provides three types of scalabilities. Spatial, temporal and quality scalabilities allow a video stream to adapt the dimensions, the number of frames per second and the fidelity depending on the target's requirements. Bit rate control is an important part of the encoding process as it allows to adapt the bitrate at the output of the encoder to the target's constraints. Unlike most of the existing approaches, our rate control technic is designed to have a low computational complexity. First, a bit rate per second constraint is specified for each video layer. Then, the available bits are dispatched among frames so that constant quality can be achieved in the decoded stream. The respect of the specified budget is finally enforced by a bit rate model, based on a simple and effective framework called $\rho$-domain. Using the statistics of the frame, this model allows us to choose the optimal quantization parameter while maintaining the amount of calculations very low. Based on this simple technic, we propose two approaches. In the first approach, each image is pre-encoded to provide the bit rate model with the statistics of the frame before rate control. This approach allows us to regulate the bit rate with great accuracy, as the bit rate error is below 3\% of the specified constraint on each type of scalability. In the second approach, rate control is performed using the information from the previous frame, so that no pre-encoding step is required and each frame is encoded only once. Thus, the impact on the encoding process is greatly reduced, at the cost of a small error increase. Finally, perceptual quality is considered when dispatching the bits among frames, so that quality variations are reduced and visual experience is improved. Our bit rate control technic for MPEG-4 SVC shows great accuracy on all spatial, temporal and quality scalabilities. Its low complexity, together with quality variation control make it a valuable contribution, especially in practical applications, for which time ressources are limited and the user-felt quality is important.
Complete list of metadatas

Cited literature [79 references]  Display  Hide  Download
Contributor : Marie Babel <>
Submitted on : Friday, March 5, 2010 - 11:03:50 AM
Last modification on : Monday, October 5, 2020 - 9:50:17 AM
Long-term archiving on: : Friday, June 18, 2010 - 10:20:10 PM


  • HAL Id : tel-00461631, version 1


Yohann Pitrey. Stratégies d'encodage pour codeur vidéo scalable. Traitement du signal et de l'image [eess.SP]. INSA de Rennes, 2009. Français. ⟨tel-00461631⟩



Record views


Files downloads