Signal Processing for High-Definition Television

This thesis presents new design techniques for source coding of video signals. The main application is high-definition television (HDTV), characterized by roughly twice the vertical and horizontal resolution of current systems; aggressive source coding is required to compress the large amount of video information into the relatively narrow channels envisioned for the new service. Video source coders conventionally use block transforms with small support (typically 8x8 pixels). Such independent blocks result in a simple scheme for switching a predictor from a motion-compensated block to a purely spatial block; this is necessary to prevent the coder from wasting capacity in some situations. Subband coders of the multiresolution or wavelet type, with their more desirable localization properties, lack of "blocking" artifacts, and better match to motion-compensated prediction errors, complicate this process of switching predictors, since the blocks now overlap. This thesis presents a novel predictive coding scheme in which subband coders can combine the benefits of good representation and flexible adaptive prediction. Thesis Supervisor: Jae S. Lim Title: Professor Thesis Supervisor: Gilbert Strang Title: Professor