Sparse Measurement Matrices for Compressed-Sensing Recovery by Bayesian Approximate Message Passing

Sparse measurement matrices with very few randomly selected +1/-1 non-zero elements are designed for use with Bayesian Approximate Message Passing as a compressed sensing recovery algorithm. Simulations show that such sparse matrices, which allow for large savings in storage and computation time, can achieve a recovery performance that is as good as the benchmark given by random Gaussian matrices.