Network-adaptive prediction-compensated polyphase multiple description image coding and contrast-guided image interpolation

Abstract

In this thesis, we propose an efficient multiple description coding (MDC) system and a novel image interpolation method. For the MDC system, it consists of two thrust contributions: 1) a new polyphase MDC scheme, called the prediction-compensated polyphase MDC (PCP-MDC), for significantly improving coding efficiency, and 2) an adaptive redundancy control (ARC) scheme for yielding optimal trade-off between coding efficiency and error resilience as follows. For the first contribution, the proposed PCP-MDC begins with a simple yet novel approach to first partition each quincunx downsampled description into two sub-descriptions, called the primary sub-description (PS) and dual sub-description (DS). The PS is encoded by exploiting the H.264/AVC intra coding first, and the DS is subject to prediction coding based on the reconstructed PS. For prediction, the mode-guided directional prediction (MDP) algorithm is developed to conduct a fast and accurate prediction for the DS.