A universal software framework for the assessment of compressed sensing algorithms

The inherent differences of the Compressed Sensing (CS) imaging technique result in advantages compared to conventional imaging for scenes containing a rather low amount of relevant data. Such information ""sparsity"" enables the reconstruction of images with a small number of low-resolution measurements, focusing on the relevant information already during the measurement process. The crucial ""sparsity"" requirement of the CS framework is fulfilled in certain threat detection applications like e.g. solar blind UV missile warning. To assess, whether CS technology is applicable for threat detection, where short acquisition times and robust and reliable detection are a must, we developed a universal framework for the assessment of the performance of compressed sensing algorithms. This framework includes the most promising algorithms from our past performance assessments and a selection of image quality metrics suited for conceivable applications. The currently implemented algorithms are TVAL3, NESTA, FPC and L1-primal-dualalgorithms. Because of the structure of the realized framework, the integration of further algorithms and metrics as well as alternative image acquisition approaches is possible. The studied data sets include scenes were the approach of a threat is modelled for our single-pixel-camera setup as well as recordings done with solar blind UV- and IR-imagers that were broken down to single-pixel-measurements of varying compression.