matlab-(含教程)通过matlab对比IRLS,OMP,MOMP,SP以及CoSaMP五种压缩感知算法的信号重构性能

%Compressed Sensing / Compressive Sensing clc; clear; close all; warning off; addpath(genpath(pwd)); rng('default'); %计数器 tic l=8; c=8; S=5; % m is the number of measurement %m=20; A=5:20; D=zeros(size(A)); D2=D; Dmod=D; D1=D; %Dp5=D; Dmod3=D; Dco=D; %Monte Carlo Dsp=D; for t=1:200 rng(t); for k=1:length(D) m = A(k); [t,k] r2 = randperm(l*c);r2=r2(1:S);rr2=r2(:); u_true = zeros(l*c,1); u_true(r2)=u_true(r2)+rand(S,1); phi = (kron((fft(eye(c))).',fft(eye(l)))); r = randperm(l*c);r=r(1:m); phi = phi(r,:); b = phi * u_true(:); [d1,u_new1] = IRLSregcomp(phi,b,1,u_true(:)); [uestOMP,d2] = OMP(phi,b,u_true,S); [uestOMPmod3,dmod3] = OMPmod3(phi,b,u_true,S); [uestco,dco] = cosamp2(phi,b,u_true,S); [uestsp,dsp] = SPA(phi,S,b,u_true); Dsp(k)=Dsp(k)+dsp; Dco(k)=Dco(k)+dco; Dmod3(k)=Dmod3(k)+dmod3; D2(k)=D2(k)+d2; D1(k)=D1(k)+d1; end end; D1=D1/t; Dco=Dco/t; Dsp=Dsp/t; Dmod=Dmod/t; Dmod3=Dmod3/t; D2=D2/t; figure(1) hold off hold on plot(A,D1,'k*--','LineWidth',2) plot(A,D2,'m*--','LineWidth',2) plot(A,Dmod3,'r*--','LineWidth',2) plot(A,Dsp,'g*--','LineWidth',2) plot(A,Dco,'o--','LineWidth',2) legend('IRLS p=1','OMP','modified OMP','SP','CoSaMP') grid xlabel('Number of measurements M') ylabel('Empirical probability of reconstruction at level 1e-3') title('N=64 S=5 \Phi : undersampled DFT')