SAR raw data compression based on geometric characteristic of Gaussian curve

Metrics to evaluate compression algorithms for raw SAR data

“…The MSE is a measure of the performance of the quantiser and gives the total absolute encoding error between the uncompressed and decompressed datasets [6, 9, 22, 24]. It can be computed as The MSE is computed using the magnitude components of the respective complex SAR datasets.…”

“…In SAR processing, the phase information is used to focus the SAR image, and therefore, knowing the phase error after decompression gives an indication of the focusing error that will be present in the final image [22]. The MPE [6, 23, 24] is computed for the phase component of the data using the following equation: where is the phase component of the complex SAR data of both sets being evaluated.…”

“…Not only are platforms becoming smaller, but SAR technology is moving towards being multi‐band, multi‐polarised, having very high resolution and also having multiple operating modes. All of these factors contribute to SAR systems producing an ever‐increasing volume of data, so despite technological advances, the on‐board storage and downlink bandwidth remain limiting factors [6, 7]. Due to the limited SWAP‐C of many platforms, minimal processing can be performed on board, and thus the data need to be transferred to a ground station for processing.…”

“…Various compression algorithms have been investigated, including the block adaptive quantiser (BAQ) [8, 9] and variations thereof [6, 10, 11], vector quantisation [9, 12], transform coding [13], trellis coded quantisation [14], wavelets [15], compressive sensing [16], predictive coding [17], coding after range focusing [14], as well as combinations of more than one algorithm [18]. The de facto standard is the BAQ [6, 19] which utilises a Lloyd‐Max quantiser [20, 21] to exploit the fact that the in‐phase (I) and quadrature (Q) components can be assumed to follow Gaussian distributions with zero mean [8, 9]. More advanced variations of this method have been implemented on new SAR systems as processing capabilities improve [7], but the underlying BAQ algorithm is still being used.…”

“…These metrics consider how significantly the data rate is reduced, the loss of information and the processing required for compression. Furthermore, the metrics evaluated consider both the data domain, where compression and decompression are performed, and the image domain, where the final SAR output is produced [6, 10, 22].…”