Robust video coder solution for wireless streaming: applications in Gaussian channels

• Autor(a) principal: Arrifano, Ângelo Miguel
• Data de Publicação: 2010
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.6/3720
• topic: Multiple description coding; Wireless channel; Noisy channel; Gaussian channel; Subband coding; Lagrangian allocation

With the technological progress in wireless communications seen in the past decade, the miniaturization of personal computers was imminent. Due to the limited availability of resources in these small devices, it has been preferable to stream the media over widely deployed networks like the Internet. However, the conventional protocols used in physical and data-link layers are not adequate for reliable video streaming over noisy wireless channels. There are several popular and well-studied mechanisms for addressing this problem, one of them being Multiple-Description-Coding. However, proposed solutions are too specialized, focusing the coding of either motion or spatial information; thus failing to address the whole problem, that is - the robust video coding. In this thesis a novel MDC video coder is presented, which was developed during an internship at the I3S laboratory - France. The full coding scheme is capable of robust transmission of Motion-Vectors and wavelet-subband information over noisy wireless channels. The former is accomplished by using a MAP-based MD-decoding algorithm available in literature, while the robust transmission of wavelet-subbands is achieved using a state-of-the-art registry-based JPEG-2000 MDC. In order to e ciently balance MV information between multiple descriptions, a novel R/D-optimizing MD bitallocation scheme is presented. As it is also important to e ciently distribute bits between subband and motion information, a global subband/motion-vector bit-allocation technique found in literature was adopted and improved. Indeed, this thesis would not be complete without the presentation of produced streams as well as of a set of backing scienti c results.