Abstract
HEVC (High Efficiency Video Coding), the latest and most popular video coding standard, has succeeded in significantly improving performance in terms of the compression efficiency by exploiting more and more spatial and temporal dependencies. Consequently, HEVC coded videos are much more susceptible to network impairments, which unless mitigated by some means, can have a significant detrimental impact on the end user’s perceived Quality of Experience (QoE). Furthermore, emerging HEVC based interactive video applications pose many challenges which cannot be resolved using traditional resource allocation algorithms or error resilience schemes. These include dynamic network conditions, the need for maintaining the QoE across a group of interacting users, strict delay constraints, as well as the geographic and temporal variations in the interactive content processing requirements, all of which must collectively determine and adapt the appropriate video coding parameters and network level resource allocation to enhance the overall QoE. To this end, this research proposes application layer and network layer methodologies to improve the end user QoE. First, an end user's video quality prediction model is proposed for HEVC based video communications under error prone channels. The model which incorporates the errors in the motion vectors, the reference pixels and the pixel clipping operations during the modelling phase, has subsequently demonstrated only a 3% prediction error whereas the state-of-the-art methods couldn't reach below 17.5% under identical conditions. Later, these distortion values are re-used inside the HEVC Rate-Distortion Optimization (RDO) process to realize a 20%-40% improvement in the BD-rate compared to the state-of-the-art error resilient schemes. Secondly, a novel motion vector estimation algorithm is proposed to select motion vectors which unlike the existing methods not only mitigate the error propagation from the previous frames but also improve the concealment accuracy of the future encoding frames. The proposed algorithm has consequently demonstrated 1.48 dB PSNR gain compared to the existing methods for the same bit rate. Finally a joint network and computational resource allocation scheme and an optimal transmission route selection algorithm is proposed to maximize the end user QoE of an interactive video application. The results demonstrates that the proposed resource allocation scheme can outperform the existing methods by a 50% margin in the serving probabilities and several orders of magnitude reduction in the computational times.