www.gusucode.com > 压缩感知的几种重构算法比较的源代码 > 压缩感知的几种重构算法比较的源代码/MMA14/cosamp2.m

    function [Sest,d]=cosamp2(Phi,u,utrue,K)

% Cosamp algorithm
%   Input
%       K : sparsity of Sest
%       Phi : measurement matrix
%       u: measured vector
%       tol1 : tolerance for approximation between successive solutions. 
%   Output
%       Sest: Solution found by the algorithm
%       d   : success index (d=1 is success, d = 0 is no convergence)
%
% Algorithm as described in "CoSaMP: Iterative signal recovery from 
% incomplete and inaccurate samples" by Deanna Needell and Joel Tropp.
% 
% This implementation was written by David Mary
%
% This script/program is released under the Commons Creative Licence
% with Attribution Non-commercial Share Alike (by-nc-sa)
% http://creativecommons.org/licenses/by-nc-sa/3.0/
% Short Disclaimer: this script is for educational purpose only.
% Longer Disclaimer see  http://igorcarron.googlepages.com/disclaimer

% Initialization
Sest=zeros(size(utrue));%不明白不知道utrue是什么
v=u;
t=1; T2=[];
while t < 101 
[k,z]=sort(abs(Phi'*v));k=flipud(k);z=flipud(z);%
Omega=z(1:2*K);%取了前2K个。
T=sort(union(Omega,T2));phit=Phi(:,T);%T为支撑集合并后的集合，PHIT为观察矩阵phi只取下标为T的列后所组成的矩阵
%keyboard
b=abs(pinv(phit)*u);%通过最小二乘法进行信号估计，pinv表示伪逆
[k3,z3]=sort((b));k3=flipud(k3);z3=flipud(z3);%把得到的信号的新的估计值进行从大到小的排序
Sest=zeros(size(utrue));
Sest(T(z3(1:K)))=abs(b(z3(1:K)));  %信号估计后进行删减，只取了前K个，并取绝对值后赋给sest,sest此时为信号估计
[k2,z2]=sort(abs(Sest));k2=flipud(k2);z2=flipud(z2);
T2=z2(1:K);%T2为信号估计值的最大K个下标集合。
v=u-Phi*Sest;%残差进行更新
d=0;n2=norm(abs(Sest-utrue),'inf');
if n2 <1e-3
    d=1;t=1e10;
    disp('CoSaMP: success');
end
 t=t+1;
 end
 if d==0
     disp('CoSaMP: failed')
 end