Context-based Entropy Coding with Space- Frequency Segmentation in Ultrasound Image Compression

Entropy coding provides for the lossless compression of data symbols and is a critical component in signal compression algorithms. In the case where there is statistical dependence between the symbols produced by the source, the performance of an entropy coder can be significantly improved by having the coder dynamically adapt to the current "context". This thesis considers the specific case of the Space-Frequency Segmentation (SFS) compression algorithm, a method that has been successfully used in the compression of ultrasound images, but which assumes essentially independent symbols in the design of its entropy coder. In particular, we focus on different approaches to picking the context set to use for the entropy coder, as well as the tradeoffs between the number of different contexts to use and the difficulty in reliably estimating the symbol statistics. These issues are addressed for both natural and ultrasound images to determine the effect of the image type.