Figure 3.

(a) Recovery rate curves for the one-dimensional BP problem. The solution with a chaotic sensing matrix is given by the blue solid line and the optimal solution given by a Gaussian random sensing matrix is in red dashed line. Curves are obtained over a 1000 trials to compute the probability estimate. (b) Example of perfect reconstruction with unitary probability for low sparsity signal (K = 10). (c) Example of imperfect reconstruction occuring with probability 1/2 for signal with a higher sparsity level (K = 20). Original and reconstructed signals are represented by a blue solid and red dashed lines, respectively.