Metrics to evaluate compressions algorithms for RAW SAR data

Modern synthetic aperture radar (SAR) systems have size, weight, power and cost (SWAP-C) limitations since platforms are becoming smaller, while SAR operating modes are becoming more complex. Due to the computational complexity of the SAR processing required for modern SAR systems, performing the processing on board the platform is not a feasible option. Thus, SAR systems are producing an ever-increasing volume of data that needs to be transmitted to a ground station for processing. Compression algorithms are utilised to reduce the data volume of the raw data. However, these algorithms can cause degradation and losses that may degrade the effectiveness of the SAR mission. This study addresses the lack of standardised quantitative performance metrics to objectively quantify the performance of SAR data-compression algorithms. Therefore, metrics were established in two different domains, namely the data domain and the image domain. The data-domain metrics are used to determine the performance of the quantisation and the associated losses or errors it induces in the raw data samples. The image-domain metrics evaluate the quality of the SAR image after SAR processing has been performed. In this study three well-known SAR compression algorithms were implemented and applied to three real SAR data sets that were obtained from a prototype airborne SAR system. The performance of these algorithms were evaluated using the proposed metrics. Important metrics in the data domain were found to be the compression ratio, the entropy, statistical parameters like the skewness and kurtosis to measure the deviation from the original distributions of the uncompressed data, and the dynamic range. The data histograms are an important visual representation of the effects of the compression algorithm on the data. An important error measure in the data domain is the signal-to-quantisation-noise ratio (SQNR), and the phase error for applications where phase information is required to produce the output. Important metrics in the image domain include the dynamic range, the impulse response function, the image contrast, as well as the error measure, signal-to-distortion-noise ratio (SDNR). The metrics suggested that all three algorithms performed well and are thus well suited for the compression of raw SAR data. The fast Fourier transform block adaptive quantiser (FFT-BAQ) algorithm had the overall best performance, but the analysis of the computational complexity of its compression steps, indicated that it is has the highest level of complexity compared to the other two algorithms. Since different levels of degradation are acceptable for different SAR applications, a trade-off can be made between the data reduction and the degradation caused by the algorithm. Due to SWAP-C limitations, there also remains a trade-off between the performance and the computational complexity of the compression algorithm.Dissertation (MEng)--University of Pretoria, 2019.Electrical, Electronic and Computer EngineeringMEngUnrestricte