Abstract
Video communication is a popular topic for research in the modem communication field. It is widely used in mobile television broadcasting, teleconferencing, gaming applications and online streaming. However, due to the unpredictable nature of transmission network, a bit stream transmitted through a noisy channel may suffer heavy loss of data. This hinders user satisfaction, through artefacts created by lost data fragments. Hence, error resilient techniques are introduced to the transmitted data stream to improve the error robustness. The research presented here, is mainly focused on the error resilient tool called redundant coding and the possible adaptability of this tool to channel conditions. Also, this thesis discusses the error resilience tools of both 2D and 3D video coding paradigms, focusing on the bottlenecks associated with the existing techniques. One of the key objectives of this research is to find an efficient trade-off among the redundant information present, bandwidth utilisation and the capacity of error robustness. An unequal error protection (UEP) based redundant motion information coding method is outlined. The prioritisation achieved through UEP helps to mitigate the amount of redundant information present while maintaining a good primary picture quality. Also, the possibility of replacing the residual information with turbo coded parity bits is discussed. This particular proposal employs the unidirectional distributed video coding architecture to generate Wyner-Ziv parity bits to improve the motion compensated frame of the redundant stream. Another section of the thesis presents a new coding architecture where encoder based H.264/AVC compatible motion compensated frames and Wyner-Ziv parity bits are used to generate a robust primary data stream, which has the capacity to withstand a large number of bit errors in WiMAX environments, specifically for low motion sequences. Finally, in the last section of the thesis, error resilience tools are investigated for 3D video coding. Much needed redundant data transmission based error resilient tools for multi-view coding are presented. The two techniques described, employ the disparity information and region of interest coding to generate the redundant information respectively to achieve the target.