PhD thesis

AuthorZouhair M. Belkoura
TitleAnalysis and Application of Turbo Coder based Distributed Video Coding
TutorProf. Dr.-Ing. Thomas Sikora
AbstractThis work investigates a Distributed Video Coding (DVC) method for single view video sources. DVC is different from hybrid block-based video coding as there is no temporal prediction included in the encoding process. This makes video codecs with low encoder workload possible. DVC in this thesis is an extension of source coding with side information, a subset of Slepian-Wolf coding. The theoretical bounds are given for source coding with systematic side information (SCSSI), a special case using systematic channel codes to perform data compression. Practical SCSSI coding using systematic turbo codes is compared to the theoretical bounds. The Berlin Distributed Video Coder (BDVC) is implemented using turbo coder based SCSSI. Potential limiting factors that bound the BDVC performance away from hybrid coders are summarised. Experimental analyses show the influence of turbo coder variation on compression performance. A review of the error resilience properties of the BDVC is given. Based on the analyses, modifications to the BDVC are proposed to improve the compression performance. Experiments show these modifications to allow rate reductions of up to 60%. Finally, a doubly hybrid video coder (DHVC) combining H.264 and BDVC is proposed. The DHVC allows to allocate workload between coder and decoder depending on the desired operating conditions.
Key wordsmultimedia signal compression, source coding, MP3, H.264, DSC, DVC