Anitori, Laura; Otten, Matern; Hoogeboom, Peter (TNO, The Hague, The Netherlands)
Maleki, Arian; Baraniuk, Richard (Department of Electrical and Computer Engineering, Rice University, Houston, USA)
In this paper we present results on application of Compressive Sensing (CS) to high resolution radar imaging and propose the adaptive Complex Approximate Message Passing (CAMP) algorithm for image reconstruction. CS provides a theoretical framework that guarantees, under certain assumptions, reconstruction of sparse signals from many fewer measurements than required by the Nyquist-Shannon sampling theorem. However, whereas most conventional imaging techniques are based on linear filtering, in CS the image is obtained from a subsampled set of measurements by means of a non-linear reconstruction algorithm. A variety of such algorithms have been proposed, and, for a given problem instance, the solution will depend on a threshold that has either to be provided by the user or estimated from the compressed measurements. In this paper, we present an adaptive version of CAMP, where the threshold is estimated from the data itself to provide a solution with minimum reconstruction error. Our results show that the adaptive CAMP algorithm can reconstruct the image with a Mean Squared Error (MSE) comparable to the reconstruction error achieved using an optimally tuned algorithm.