Sparse Signal Reconstruction of Inverse Synthetic Aperture Radar based on Intra-Cluster Correlation

In many practical applications such as inverse synthetic aperture radar (ISAR), signals have block structure. That is the non-zero entries appear in a number of block (cluster). In these signals, the prior information about the number, size, and location of non-zero blocks is not available. According to continuity pattern of ISAR signal, a neighborhood of the coefficients can be considered for dominant scattering. In this paper, we propose a method to reconstruct ISAR sparse signal based on Bayesian compressive sensing according to intra-cluster correlation between neighboring of dominant scattering points. The sparsity of dominant scattering point not only depended on its hyperparameter, but also on neighboring hyperparameters. The proposed method does not depend on prior knowledge of the block structure. In order to model intra-cluster correlations between neighboring scattering points, a Gaussian prior distribution is used and finally variational Bayesian (VB) inference has been used to learn the hyperparameters and sparse signal estimation. The simulation results show the superiority of the proposed method compared with other methods in terms of the correlation, recovery error, perfect recovery percentage, entropy and image contrast.