Matlab单像素成像算法比较

Code demo for single-pixel imaging (SPI) with different reconstruction methods including [1] differential ghost imaging (DGI) [2] gradient descent (GD) [3] conjugate gradient descent (CGD) [4] Poisson maximum likelihood (Poisson) [5] alternating projection (AP) [6] sparse representation compressive sensing (Sparse) [7] total variation compressive sensing (TV)

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% By Liheng Bian June 22 2017. Contact me: lihengbian@gmail.com.
% This demo does the simulation of single-pixel imaging with different reconstruction methods.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

clear; 
clc;
close all;
addpath（genpath（pwd））;

%% Parameters
num_pixel = 64; % number of image pixels in each dimension
samplingRatio = 0.5;

para.tol = 1e-2;
para.min_iter = 30;
para.x0flag = 0; % initialization flag of the reconstructed image 0: all one; 1: pinv（A）*b.

%% Measurements
im = im2double（imread（‘cameraman.tif‘））;
im = imresize（im[num_pixelnum_pixel]）;
figureimshow（im[]‘InitialMagnification‘1000）; title（‘Scene image‘）;

num_pattern = round（samplingRatio * num_pixel * num_pixel）; % number of illumination patterns
patterns =  rand（num_pixelnum_pixelnum_pattern）;
measurements = sum（sum（repmat（im[11num_pattern]） .* patterns））;
measurements = reshape（measurements[]1）;

%% initialization
[row col m] = size（patterns）;
P = reshape（patterns [row*col m]）;
P = P‘; % each row represents a pattern

if para.x0flag == 1
    para.x0 = pinv（P）*measurements;
else
    para.x0 = ones（row * col1）;
end

%% 1. SPI differential ghost imaging （DGI） reconstruction
ind = 1;
fprintf（‘Begin reconstruction of DGI. \n‘）;
tic
[im_r_DGI] = fun_SPI_R_DGI（patterns measurements）;
runTime（ind2） = toc;
figureimshow（im_r_DGI[]‘InitialMagnification‘1000）; title（‘Recovered im using DGI method‘）;
im_r（::ind） = im_r_DGI;

%% 2. SPI gradient descent （GD） reconstruction
ind = 2;
fprintf（‘Begin reconstruction of GD. \n‘）;
tic
[im_r_GD totaliter] = fun_SPI_R_GD（patterns measurements para）;
runTime（ind2） = toc;
runTime（ind1） = totaliter;
figureimshow（im_r_GD[]‘InitialMagnification‘1000）; title（‘Recovered im using GD method‘）;
im_r（::ind） = im_r_GD;

%% 3. SPI conjugate gradient descent （CGD） reconstruction
ind = 3;
fprintf（‘Begin reconstruction of CGD. \n‘）;
tic
[im_r_CGD totaliter] = fun_SPI_R_CGD（patterns measurements para）;
runTime（ind2） = toc;
runTime（ind1） = totaliter;
figureimshow（im_r_CGD[]‘InitialMagnification‘1000）; title（‘Recovered im using CGD method‘）;
im_r（::ind） = im_r_CGD;

%% 4. SPI Poisson maximum likelihood reconstruction
ind = 4;
fprintf（‘Begin reconstruction of Poisson. \n‘）;
tic
[im_r_Poisson totaliter] = fun_SPI_R_Poisson（patterns measurements para）;
runTime（ind2） = toc;
runTime（ind1） = totaliter;
figureimshow（im_r_Poisson[]‘InitialMagnification‘1000）; title（‘Recovered im using Poisson method‘）;
im_r（::ind） = im_r_Poisson;

%% 5. SPI alternating projection （AP） reconstruction
ind = 5;
fprintf（‘Begin reconstruction of AP. \n‘）;
tic
[im_r_AP totaliter] = fun_SPI_R_AP（patterns measurements para）;
runTime（ind2） = toc;
runTime（ind1） = totaliter;
figureimshow（im_r_AP[]‘InitialMagnification‘1000）; title（‘Recovered im using AP method‘）;
im_r（::ind） = im_r_AP;

%% 6. SPI Sparse representa

 属性            大小     日期    时间   名称
----------- ---------  ---------- -----  ----
     文件        4714  2017-06-22 17:46  Demo.m
     文件         806  2017-10-24 15:52  fun_error.m
     文件        2042  2017-11-21 16:05  fun_SPI_R_AP.m
     文件        1949  2017-11-21 16:11  fun_SPI_R_CGD.m
     文件        1414  2016-06-29 06:22  fun_SPI_R_DGI.m
     文件        2045  2017-06-22 10:10  fun_SPI_R_GD.m
     文件        2648  2017-06-22 10:10  fun_SPI_R_Poisson.m
     文件        4674  2017-10-21 17:30  fun_SPI_R_Sparse.m
     文件        5691  2017-10-24 15:54  fun_SPI_R_TV.m
     文件        4655  2017-10-24 15:54  README.txt