Fast lp-sparse Bayesian learning for compressive sensing reconstruction

The problem of compressive sensing reconstruction can be come down to the problem of solving a sparse linear model. In this paper, the l_p(0<;p≤1)-type prior that encourages sparsity of the weights is used to the sparse linear model. Motivated by maximum a posteriori estimation and sparse Bayesian learning, a stage-wise fast l_p-sparse Bayesian learning (SF-l_p-SBL) algorithm is proposed through integrating with a fast sequential learning scheme and a stage-wise strategy. The experiments demonstrate that SF-l_p-SBL is a fast and effective CS reconstruction algorithm.