Range-azimuth SAR imaging based on Bayesian Compressive Sensing

In this paper we develop an efficient anti-noise imaging algorithm based on Bayesian Compressive Sensing (BCS) theory. Random sampling is applied in range and azimuth direction respectively, and sparse dictionary matrixes are designed independently in each direction according to imaging geometry model. At last, BCS theory is used to reconstruct SAR image. BCS theory takes the prior knowledge of targets into consideration as well as the additional Gaussian noise, and therefore it can reconstruct images more clearly and robustly. The computer experiments show that the imaging algorithm developed in this paper performs better in anti-noise ability and spatial resolution compared to FFT-based algorithms and the orthogonal matching pursuit (OMP) algorithm. Thus we can obtain higher image resolution. When compared to the algorithms where only one sparse dictionary is designed for two-dimension area, this algorithm has much lower computational complexity and needs less memory.