Scalable Joint Source-Channel Coding of Image and Video Signals

The generation of scalable encoded representations for image and video signals is considered nowadays to be essential as it allows for accommodating one single encoded stream to the available bandwidth of the transmission channel, and to a plethora of display devices presenting different needs in terms of resolution, frame rate and quality. Additionally, when transmitting the scalable encoded streams over lossy transmission channels, resilience against losses is of critical importance. Driven by these requirements, this thesis focuses on the design and analysis of reliable coding and transmission approaches for scalable image and video representations transmitted over error-prone networks. The first contribution of this dissertation consists in the design of a generic scalable Joint Source-Channel Coding (JSCC) methodology that minimizes the end-to-end distortion in the transmission of scalable image and video streams over bandwidth-limited lossy, memoryless channels. Error protection within this framework is attained using Forward Error Correcting (FEC) codes. The second contribution of this dissertation is a comparative analysis of scalable JSCC and scalable Multiple Description Coding (MDC), as two representative competing approaches that simultaneously provide scalability and error resilient coding. Finally, the third contribution of this dissertation consists in the development of a hybrid MDC-JSCC methodology that allows for the transmission of multiple descriptions over one or multiple channels. One concludes that this dissertation proposes novel scalable error-resilient coding approaches that enable transmission of image and video data over error prone channels, providing competitive performance against state-of-the-art techniques.