A new technique for low bit rate scalar quantization of image sub-band coefficients

A new method for low bit rate Scalar Quantization of image sub-band coefficients has been identified. This method is based on a phase dispersion technique originally developed for the design of pulse compression radar waveforms. The technique forms the basis for a lossless, invertable transform that converts the coefficients to an intermediate domain. Components in this domain have the property that they are approximately independent and identically distributed. A single quantizer and entropy coder can efficiently be applied to all the components. The increased efficiency and reduced overhead enables the design of an image encoder with increased performance over currently reported methods especially at low bit rates. The structure of the transform results in the uniform distribution of quantization noise across the image and the image spectrum. This results in reconstructed images with higher subjective quality and minimized artifacts over other standard techniques.; The project proceeded as follows. First, a closed form model of the phase dispersion technique was developed. The properties of the closed form model were examined. Next, a complete image compression system was designed based on the phase dispersion model. A complete encoder and decoder was designed and simulated using the MATLAB Version 4.2 tool set. Simulation experiments were run to assess the Rate vs. Distortion performance and computational requirements of the encoder and decoder. A standard database of monochrome test images was used. The performance results that were obtained were used for relevant Rate vs. Distortion comparison with the benchmark techniques.; Lastly, the performance of the method in a set of real world communications applications was examined. Experiments were run for three HF Communications systems, an HF Forward Transmission Communications System, an HF ARQ Communications System and an HF Hybrid ARQ/Forward Transmission System.