# HG changeset patch # User idamnjanovic # Date 1300117313 0 # Node ID 623fcf3a69b1cfd0c8b2bd075881c4ed98664e48 # Parent 83de4ea524df58516b5202b81bf1ffebad30ec4a diff -r 83de4ea524df -r 623fcf3a69b1 examples/Image Denoising/SMALL_ImgDenoise_DL_test_KSVDvsRLSDLA.m --- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/examples/Image Denoising/SMALL_ImgDenoise_DL_test_KSVDvsRLSDLA.m Mon Mar 14 15:41:53 2011 +0000 @@ -0,0 +1,377 @@ +%% DICTIONARY LEARNING FOR IMAGE DENOISING +% This file contains an example of how SMALLbox can be used to test different +% dictionary learning techniques in Image Denoising problem. +% It calls generateImageDenoiseProblem that will let you to choose image, +% add noise and use noisy image to generate training set for dictionary +% learning. +% Three dictionary learning techniques were compared: +% - KSVD - M. Elad, R. Rubinstein, and M. Zibulevsky, "Efficient +% Implementation of the K-SVD Algorithm using Batch Orthogonal +% Matching Pursuit", Technical Report - CS, Technion, April 2008. +% - KSVDS - R. Rubinstein, M. Zibulevsky, and M. Elad, "Learning Sparse +% Dictionaries for Sparse Signal Approximation", Technical +% Report - CS, Technion, June 2009. +% - SPAMS - J. Mairal, F. Bach, J. Ponce and G. Sapiro. Online +% Dictionary Learning for Sparse Coding. International +% Conference on Machine Learning,Montreal, Canada, 2009 +% +% +% Ivan Damnjanovic 2010 +%% + + + +% If you want to load the image outside of generateImageDenoiseProblem +% function uncomment following lines. This can be useful if you want to +% denoise more then one image for example. +clear; +TMPpath=pwd; +FS=filesep; +[pathstr1, name, ext, versn] = fileparts(which('SMALLboxSetup.m')); +cd([pathstr1,FS,'data',FS,'images']); +load('test_image.mat'); +% [filename,pathname] = uigetfile({'*.png;'},'Select a file containin pre-calculated notes'); +% [pathstr, name, ext, versn] = fileparts(filename); +% test_image = imread(filename); +% test_image = double(test_image); +% cd(TMPpath); +% SMALL.Problem.name=name; + +noise_level=[10 20 25 50 100]; +% Defining Image Denoising Problem as Dictionary Learning +% Problem. As an input we set the number of training patches. +for noise_ind=1:1 +for im_num=4:4 +SMALL.Problem = generateImageDenoiseProblem(test_image(im_num).i, 40000, '',512, noise_level(noise_ind)); +SMALL.Problem.name=im_num; + +results(noise_ind,im_num).noisy_psnr=SMALL.Problem.noisy_psnr; + +%% +% Use KSVD Dictionary Learning Algorithm to Learn overcomplete dictionary + +% Initialising Dictionary structure +% Setting Dictionary structure fields (toolbox, name, param, D and time) +% to zero values + +SMALL.DL(1)=SMALL_init_DL(); + +% Defining the parameters needed for dictionary learning + +SMALL.DL(1).toolbox = 'KSVD'; +SMALL.DL(1).name = 'ksvd'; + +% Defining the parameters for KSVD +% In this example we are learning 256 atoms in 20 iterations, so that +% every patch in the training set can be represented with target error in +% L2-norm (EData) +% Type help ksvd in MATLAB prompt for more options. + +Edata=sqrt(prod(SMALL.Problem.blocksize)) * SMALL.Problem.sigma * SMALL.Problem.gain; +maxatoms = floor(prod(SMALL.Problem.blocksize)/2); +SMALL.DL(1).param=struct(... + 'Edata', Edata,... + 'initdict', SMALL.Problem.initdict,... + 'dictsize', SMALL.Problem.p,... + 'exact', 1, ... + 'iternum', 20,... + 'memusage', 'high'); + +% Learn the dictionary + +SMALL.DL(1) = SMALL_learn(SMALL.Problem, SMALL.DL(1)); + +% Set SMALL.Problem.A dictionary +% (backward compatiblity with SPARCO: solver structure communicate +% only with Problem structure, ie no direct communication between DL and +% solver structures) + +SMALL.Problem.A = SMALL.DL(1).D; +SMALL.Problem.reconstruct = @(x) ImgDenoise_reconstruct(x, SMALL.Problem); + +%% +% Initialising solver structure +% Setting solver structure fields (toolbox, name, param, solution, +% reconstructed and time) to zero values + +SMALL.solver(1)=SMALL_init_solver; + +% Defining the parameters needed for image denoising + +SMALL.solver(1).toolbox='ompbox'; +SMALL.solver(1).name='omp2'; +SMALL.solver(1).param=struct(... + 'epsilon',Edata,... + 'maxatoms', maxatoms); + +% Denoising the image - SMALL_denoise function is similar to SMALL_solve, +% but backward compatible with KSVD definition of denoising + +SMALL.solver(1)=SMALL_solve(SMALL.Problem, SMALL.solver(1)); +SMALL.solver(1).reconstructed.psnr +%% +% Use KSVDS Dictionary Learning Algorithm to denoise image + +% Initialising solver structure +% Setting solver structure fields (toolbox, name, param, solution, +% reconstructed and time) to zero values +% +% SMALL.DL(2)=SMALL_init_DL(); +% +% % Defining the parameters needed for dictionary learning +% +% SMALL.DL(2).toolbox = 'KSVDS'; +% SMALL.DL(2).name = 'ksvds'; +% +% % Defining the parameters for KSVDS +% % In this example we are learning 256 atoms in 20 iterations, so that +% % every patch in the training set can be represented with target error in +% % L2-norm (EDataS). We also impose "double sparsity" - dictionary itself +% % has to be sparse in the given base dictionary (Tdict - number of +% % nonzero elements per atom). +% % Type help ksvds in MATLAB prompt for more options. +% +% +% SMALL.DL(2).param=struct(... +% 'Edata', Edata, ... +% 'Tdict', 6,... +% 'stepsize', 1,... +% 'dictsize', SMALL.Problem.p,... +% 'iternum', 20,... +% 'memusage', 'high'); +% SMALL.DL(2).param.initA = speye(SMALL.Problem.p); +% SMALL.DL(2).param.basedict{1} = odctdict(8,16); +% SMALL.DL(2).param.basedict{2} = odctdict(8,16); +% +% % Learn the dictionary +% +% SMALL.DL(2) = SMALL_learn(SMALL.Problem, SMALL.DL(2)); + +% Set SMALL.Problem.A dictionary and SMALL.Problem.basedictionary +% (backward compatiblity with SPARCO: solver structure communicate +% only with Problem structure, ie no direct communication between DL and +% solver structures) + +SMALL.Problem.A = SMALL.Problem.initdict; +% SMALL.Problem.basedict{1} = SMALL.DL(2).param.basedict{1}; +% SMALL.Problem.basedict{2} = SMALL.DL(2).param.basedict{2}; +SMALL.DL(2).D=SMALL.Problem.initdict; +SparseDict=0; +SMALL.Problem.reconstruct = @(x) ImgDenoise_reconstruct(x, SMALL.Problem, SparseDict); + +%% +% Initialising solver structure +% Setting solver structure fields (toolbox, name, param, solution, +% reconstructed and time) to zero values + +SMALL.solver(2)=SMALL_init_solver; + +% Defining the parameters needed for image denoising + +SMALL.solver(2).toolbox='ompbox'; +SMALL.solver(2).name='omp2'; +SMALL.solver(2).param=struct(... + 'epsilon',Edata,... + 'maxatoms', maxatoms); + +% Denoising the image - SMALL_denoise function is similar to SMALL_solve, +% but backward compatible with KSVD definition of denoising +% Pay attention that since implicit base dictionary is used, denoising +% can be much faster then using explicit dictionary in KSVD example. + +SMALL.solver(2)=SMALL_solve(SMALL.Problem, SMALL.solver(2)); +%% + +for i =1:1 + +X=SMALL.Problem.b1; +X_norm=sqrt(sum(X.^2, 1)); +[X_norm_sort, p]=sort(X_norm); +p1=p(X_norm_sort>Edata); +if size(p1,2)>140000 + p2 = randperm(size(p1,2)); + p2=sort(p2(1:40000)); + size(p2,2) + SMALL.Problem.b=X(:,p1(p2)); +else + size(p1,2) + SMALL.Problem.b=X(:,p1); + +end + +lambda=0.9998 + +% Use Recursive Least Squares +% to Learn overcomplete dictionary + +% Initialising Dictionary structure +% Setting Dictionary structure fields (toolbox, name, param, D and time) +% to zero values + +SMALL.DL(3)=SMALL_init_DL(); + +% Defining fields needed for dictionary learning + +SMALL.DL(3).toolbox = 'SMALL'; +SMALL.DL(3).name = 'SMALL_rlsdla'; +SMALL.DL(3).param=struct(... + 'Edata', Edata,... + 'initdict', SMALL.Problem.initdict,... + 'dictsize', SMALL.Problem.p,... + 'forgettingMode', 'FIX',... + 'forgettingFactor', lambda); + +% % Type 'help mexTrainDL in MATLAB prompt for explanation of parameters. +% +% SMALL.DL(3).param=struct(... +% 'D', SMALL.Problem.initdict,... +% 'K', SMALL.Problem.p,... +% 'lambda', 2,... +% 'iter', 200,... +% 'mode', 3, ... +% 'modeD', 0); + +% Learn the dictionary + +SMALL.DL(3) = SMALL_learn(SMALL.Problem, SMALL.DL(3)); +%SMALL.DL(3).D(:,1)=SMALL.DL(1).D(:,1); +% +% % Set SMALL.Problem.A dictionary +% % (backward compatiblity with SPARCO: solver structure communicate +% % only with Problem structure, ie no direct communication between DL and +% % solver structures) +% +% +% +% %% +% % Initialising solver structure +% % Setting solver structure fields (toolbox, name, param, solution, +% % reconstructed and time) to zero values +% SMALL.Problem.A = SMALL.DL(1).D; +% SMALL.Problem.reconstruct = @(x) ImgDenoise_reconstruct(x, SMALL.Problem); +% maxatoms=5; +% SMALL.solver(3)=SMALL_init_solver; +% +% % Defining the parameters needed for denoising +% +% % SMALL.solver(3).toolbox='SPAMS'; +% % SMALL.solver(3).name='mexLasso'; +% % SMALL.solver(3).param=struct(... +% % 'mode', 1, ... +% % 'lambda',Edata*Edata,... +% % 'L', maxatoms); +% % % Denoising the image - SMALL_denoise function is similar to SMALL_solve, +% % % but backward compatible with KSVD definition of denoising +% % +% % SMALL.solver(3)=SMALL_solve(SMALL.Problem, SMALL.solver(3)); +% SMALL.solver(3).toolbox='SMALL'; +% SMALL.solver(3).name='SMALL_cgp'; +% SMALL.solver(3).param=sprintf('%d, %.2f', maxatoms, sqrt(Edata)); +% % Denoising the image - SMALL_denoise function is similar to SMALL_solve, +% % but backward compatible with KSVD definition of denoising +% +% SMALL.solver(3)=SMALL_solve(SMALL.Problem, SMALL.solver(3)); + +% %% +% % Use RLS-DLA +% +% % Initialising Dictionary structure +% % Setting Dictionary structure fields (toolbox, name, param, D and time) +% % to zero values +% +% SMALL.DL(3)=SMALL_init_DL(); +% +% % Defining fields needed for dictionary learning +% +% SMALL.DL(3).toolbox = 'mpv2'; +% SMALL.DL(3).name = 'rlsdla'; +% +% % Type 'help mexTrainDL in MATLAB prompt for explanation of parameters. +% +% SMALL.DL(3).param=struct(... +% 'D', SMALL.Problem.initdict,... +% 'K', SMALL.Problem.p,... +% 'abs', Edata*Edata,... +% 'lambda', 0.995,... +% 'iternum',1); +% +% % Learn the dictionary +% +% SMALL.DL(3) = SMALL_learn(SMALL.Problem, SMALL.DL(3)); +% +% % Set SMALL.Problem.A dictionary +% % (backward compatiblity with SPARCO: solver structure communicate +% % only with Problem structure, ie no direct communication between DL and +% % solver structures) +% +% + +%% +% Initialising solver structure +% Setting solver structure fields (toolbox, name, param, solution, +% reconstructed and time) to zero values +%SMALL.DL(3).D(:,225:256)=0; +SMALL.Problem.A = SMALL.DL(3).D; +SMALL.Problem.reconstruct = @(x) ImgDenoise_reconstruct(x, SMALL.Problem); +%maxatoms=32; +SMALL.solver(3)=SMALL_init_solver; + +% Defining the parameters needed for denoising + +% SMALL.solver(3).toolbox='SPAMS'; +% SMALL.solver(3).name='mexLasso'; +% SMALL.solver(3).param=struct(... +% 'mode', 1, ... +% 'lambda',Edata*Edata,... +% 'L', maxatoms); +% % Denoising the image - SMALL_denoise function is similar to SMALL_solve, +% % but backward compatible with KSVD definition of denoising +% +% SMALL.solver(3)=SMALL_solve(SMALL.Problem, SMALL.solver(3)); + +% Initialising solver structure +% Setting solver structure fields (toolbox, name, param, solution, +% reconstructed and time) to zero values + +SMALL.solver(3)=SMALL_init_solver; + +% Defining the parameters needed for image denoising + +SMALL.solver(3).toolbox='ompbox'; +SMALL.solver(3).name='omp2'; +SMALL.solver(3).param=struct(... + 'epsilon',Edata,... + 'maxatoms', maxatoms); +% SMALL.solver(3).toolbox='SPAMS'; +% SMALL.solver(3).name='mexLasso'; +% SMALL.solver(3).param=struct(... +% 'mode', 2, ... +% 'lambda',40,... +% 'L', maxatoms); + +% Denoising the image - SMALL_denoise function is similar to SMALL_solve, +% but backward compatible with KSVD definition of denoising + +SMALL.solver(3)=SMALL_solve(SMALL.Problem, SMALL.solver(3)); +% Plot results and save midi files +SMALL.solver(3).reconstructed.psnr +% show results % + +SMALL_ImgDeNoiseResult(SMALL); +end +results(noise_ind,im_num).psnr.ksvd=SMALL.solver(1).reconstructed.psnr; +results(noise_ind,im_num).psnr.odct=SMALL.solver(2).reconstructed.psnr; +results(noise_ind,im_num).psnr.rlsdla=SMALL.solver(3).reconstructed.psnr; +results(noise_ind,im_num).vmrse.ksvd=SMALL.solver(1).reconstructed.vmrse; +results(noise_ind,im_num).vmrse.odct=SMALL.solver(2).reconstructed.vmrse; +results(noise_ind,im_num).vmrse.rlsdla=SMALL.solver(3).reconstructed.vmrse; +results(noise_ind,im_num).ssim.ksvd=SMALL.solver(1).reconstructed.ssim; +results(noise_ind,im_num).ssim.odct=SMALL.solver(2).reconstructed.ssim; +results(noise_ind,im_num).ssim.rlsdla=SMALL.solver(3).reconstructed.ssim; + +results(noise_ind,im_num).time.ksvd=SMALL.solver(1).time+SMALL.DL(1).time; +results(noise_ind,im_num).time.rlsdla.time=SMALL.solver(3).time+SMALL.DL(3).time; +%clear SMALL; +end +end +save results.mat results