Coding of motion picture residues without spatial transforms

A coding scheme for prediction-image residues in video coding is developed in this thesis. The strategy is to use the zerotree structure invented by Shapiro and extended by Said and Pearlman. Spatial patterns of pixels are defined to occupy successive levels of the tree structure. These patterns replace the standard quad-tree ordering of wavelet coefficient values with simple pixel values. The motivation for this is that prediction image residues have little spatial structure. Simulations to evaluate the strategy are included in the thesis. The results demonstrate that video coding with this simple zerotree pattern coding algorithm for motion picture residues without using any spatial transforms can achieve good trade-offs between compression performance and channel error resilience and robustness, and provide additional error concealment with the help of a simple post-processing technique. It allows for operation in noisy channel conditions, without the aid of error correction or detection techniques