A multi-objective performance optimisation framework for video coding

Digital video technologies have become an essential part of the way visual information is created, consumed and communicated. However, due to the unprecedented growth of digital video technologies, competition for bandwidth resources has become fierce. This has highlighted a critical need for optimising the performance of video encoders. However, there is a dual optimisation problem, wherein, the objective is to reduce the buffer and memory requirements while maintaining the quality of the encoded video. Additionally, through the analysis of existing video compression techniques, it was found that the operation of video encoders requires the optimisation of numerous decision parameters to achieve the best trade-offs between factors that affect visual quality; given the resource limitations arising from operational constraints such as memory and complexity. The research in this thesis has focused on optimising the performance of the H.264/AVC video encoder, a process that involved finding solutions for multiple conflicting objectives. As part of this research, an automated tool for optimising video compression to achieve an optimal trade-off between bit rate and visual quality, given maximum allowed memory and computational complexity constraints, within a diverse range of scene environments, has been developed. Moreover, the evaluation of this optimisation framework has highlighted the effectiveness of the developed solution.