Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.595763
Title: Efficient algorithms for scalable video coding
Author: Lu, Xin
Awarding Body: University of Warwick
Current Institution: University of Warwick
Date of Award: 2013
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
A scalable video bitstream specifically designed for the needs of various client terminals, network conditions, and user demands is much desired in current and future video transmission and storage systems. The scalable extension of the H.264/AVC standard (SVC) has been developed to satisfy the new challenges posed by heterogeneous environments, as it permits a single video stream to be decoded fully or partially with variable quality, resolution, and frame rate in order to adapt to a specific application. This thesis presents novel improved algorithms for SVC, including: 1) a fast inter-frame and inter-layer coding mode selection algorithm based on motion activity; 2) a hierarchical fast mode selection algorithm; 3) a two-part Rate Distortion (RD) model targeting the properties of different prediction modes for the SVC rate control scheme; and 4) an optimised Mean Absolute Difference (MAD) prediction model. The proposed fast inter-frame and inter-layer mode selection algorithm is based on the empirical observation that a macroblock (MB) with slow movement is more likely to be best matched by one in the same resolution layer. However, for a macroblock with fast movement, motion estimation between layers is required. Simulation results show that the algorithm can reduce the encoding time by up to 40%, with negligible degradation in RD performance. The proposed hierarchical fast mode selection scheme comprises four levels and makes full use of inter-layer, temporal and spatial correlation aswell as the texture information of each macroblock. Overall, the new technique demonstrates the same coding performance in terms of picture quality and compression ratio as that of the SVC standard, yet produces a saving in encoding time of up to 84%. Compared with state-of-the-art SVC fast mode selection algorithms, the proposed algorithm achieves a superior computational time reduction under very similar RD performance conditions. The existing SVC rate distortion model cannot accurately represent the RD properties of the prediction modes, because it is influenced by the use of inter-layer prediction. A separate RD model for inter-layer prediction coding in the enhancement layer(s) is therefore introduced. Overall, the proposed algorithms improve the average PSNR by up to 0.34dB or produce an average saving in bit rate of up to 7.78%. Furthermore, the control accuracy is maintained to within 0.07% on average. As aMADprediction error always exists and cannot be avoided, an optimisedMADprediction model for the spatial enhancement layers is proposed that considers the MAD from previous temporal frames and previous spatial frames together, to achieve a more accurateMADprediction. Simulation results indicate that the proposedMADprediction model reduces the MAD prediction error by up to 79% compared with the JVT-W043 implementation.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.595763  DOI: Not available
Keywords: QA76 Electronic computers. Computer science. Computer software ; TK Electrical engineering. Electronics Nuclear engineering
Share: