Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.489841
Title: Application of Finite State Methods to Object Based Coding and Compressed Domain Processing
Author: Abdel-Maguid, Mohamed M.
Awarding Body: University of Staffordshire
Current Institution: Staffordshire University
Date of Award: 2007
Availability of Full Text:
Access from EThOS:
Abstract:
The aim of this work is to investigate the applicability of finite state methods to coding and compressed processing in the context of object based video. On the coding scene, an object texture coding scheme based on Weighted Finite Automata (WFA) is proposed. The scheme is implemented and benchmarked against the MJPEG, MJPEG2000 and H.264 (MPEG-4/AVC part 10) coding schemes. Rate-distortion curves for standard test video sequences at both Common Intermediate Format (CIF) and Quarter CIF (QCIF) are presented. Results obtained demonstrate that object texture coding using WFA outperforms both MJPEG and MJPEG2000 in terms of rate distortion for very low bit rates and for small resolutions. This makes WFA attractive for applications that require low bitrates and use small screen sizes such as wireless mobile video applications. The results also indicate that WFA is generally at par with H.264 for low bit rates. Three object padding techniques are evaluated for the purpose of adapting WFA to arbitrary shaped objects coding. Results suggest that zero padding outperforms both repeat padding and Low Pass Extrapolation (LPE) padding when used in conj unction with WFA encoding. On the compressed domain processing scene, a new technique for object shape compressed domain processing is proposed and critically evaluated by measuring compression efficiency; results obtained demonstrate the superior performance ofthe proposed technique. Furthermore, a novel technique for processing multiple objects simultaneously in the compressed domain is introduced; the technique presents two algorithms: an object shape masks video combiner and an object textures combiner. The performance of the proposed algorithms is evaluated in terms of processing delay and overall perceptual quality degradation. The proposed compressed domain processing techniques introduce little quality degradation compared with pixel domain processing; in many cases no quality degradation is introduced. The techniques proposed in this dissertation can potentially be attractive to video lliid image applications that require very low bit rates and applications where processing platforms have limited processing power. Mobile video surveillance, mobile video conferencing and video sensor networks are examples of such applications. Supplied by The British Library - 'The world's knowledge'
Supervisor: Not available Sponsor: Not available
Qualification Name: University of Staffordshire, 2007 Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.489841  DOI: Not available
Share: