Input: Sampling matrix , noisy sample vector , sparsity level s Output: An s-sparse approximation of the target signal |
1) Initialization: {Trivial initial approximation} {Current samples = input samples}
2) repeat a) b) {Form signal proxy} c) {Identify large components} d) {Merge supports} e) {Signal estimation by least-squares} f) g) If 0.8 (norm(r)) < norm(v):
h) {Prune to obtain next approximation} i) {Update current samples} until halting criterion true |