# HG changeset patch
# User Ivan Damnjanovic lnx <ivan.damnjanovic@eecs.qmul.ac.uk>
# Date 1311939352 -3600
# Node ID af307f247ac7cabc02d8507d57cc7a818ffa69fb
# Parent  485747bf39e0de934f92586589897ac5bac3fcae
Example scripts for Two Step Dictionary Learning - Image Denoising experiments.

diff -r 485747bf39e0 -r af307f247ac7 DL/two-step DL/SMALL_two_step_DL.m
--- a/DL/two-step DL/SMALL_two_step_DL.m	Thu Jul 28 15:49:32 2011 +0100
+++ b/DL/two-step DL/SMALL_two_step_DL.m	Fri Jul 29 12:35:52 2011 +0100
@@ -85,7 +85,7 @@
 
 % show dictonary every specified number of iterations
 
-if (isfield(DL.param,'show_dict'))
+if isfield(DL.param,'show_dict')
     show_dictionary=1;
     show_iter=DL.param.show_dict;
 else
@@ -100,19 +100,22 @@
 
 tmpTraining = Problem.b1;
 Problem.b1 = sig;
-Problem = rmfield(Problem, 'reconstruct');
+if isfield(Problem,'reconstruct')
+    Problem = rmfield(Problem, 'reconstruct');
+end
 solver.profile = 0;
 
 % main loop %
 
 for i = 1:iternum
+    Problem.A = dico;
     solver = SMALL_solve(Problem, solver);
     [dico, solver.solution] = dico_update(dico, sig, solver.solution, ...
         typeUpdate, flow, learningRate);
     if (decorrelate)
         dico = dico_decorr(dico, mu, solver.solution);
     end
-    Problem.A = dico;
+    
    if ((show_dictionary)&&(mod(i,show_iter)==0))
        dictimg = SMALL_showdict(dico,[8 8],...
             round(sqrt(size(dico,2))),round(sqrt(size(dico,2))),'lines','highcontrast');  
@@ -124,4 +127,16 @@
 Problem.b1 = tmpTraining;
 DL.D = dico;
 
+end
+
+function Y = colnorms_squared(X)
+
+% compute in blocks to conserve memory
+Y = zeros(1,size(X,2));
+blocksize = 2000;
+for i = 1:blocksize:size(X,2)
+  blockids = i : min(i+blocksize-1,size(X,2));
+  Y(blockids) = sum(X(:,blockids).^2);
+end
+
 end
\ No newline at end of file
diff -r 485747bf39e0 -r af307f247ac7 examples/Image Denoising/SMALL_ImgDenoise_DL_test_KSVDvsRLSDLAvsTwoStepMOD.m
--- a/examples/Image Denoising/SMALL_ImgDenoise_DL_test_KSVDvsRLSDLAvsTwoStepMOD.m	Thu Jul 28 15:49:32 2011 +0100
+++ b/examples/Image Denoising/SMALL_ImgDenoise_DL_test_KSVDvsRLSDLAvsTwoStepMOD.m	Fri Jul 29 12:35:52 2011 +0100
@@ -47,7 +47,7 @@
 
 %   Here we loop through different noise levels and images 
 
-for noise_ind=2:2
+for noise_ind=4:4
 for im_num=1:1
 
 % Defining Image Denoising Problem as Dictionary Learning
@@ -171,7 +171,6 @@
     'initdict', SMALL.Problem.initdict,...
     'dictsize', SMALL.Problem.p,...
     'iternum', 40,...
-    'mu', 0.7,...
     'show_dict', 1);
 
 %   Learn the dictionary
diff -r 485747bf39e0 -r af307f247ac7 examples/Image Denoising/SMALL_ImgDenoise_DL_test_KSVDvsTwoStepKSVD.m
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/examples/Image Denoising/SMALL_ImgDenoise_DL_test_KSVDvsTwoStepKSVD.m	Fri Jul 29 12:35:52 2011 +0100
@@ -0,0 +1,202 @@
+%%  Dictionary Learning for Image Denoising - KSVD vs Recursive Least Squares
+%
+%   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.
+%   Two 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.
+%   -   RLS-DLA - Skretting, K.; Engan, K.; , "Recursive Least Squares
+%       Dictionary Learning Algorithm," Signal Processing, IEEE Transactions on,
+%       vol.58, no.4, pp.2121-2130, April 2010
+%
+
+
+%   Centre for Digital Music, Queen Mary, University of London.
+%   This file copyright 2011 Ivan Damnjanovic.
+%
+%   This program is free software; you can redistribute it and/or
+%   modify it under the terms of the GNU General Public License as
+%   published by the Free Software Foundation; either version 2 of the
+%   License, or (at your option) any later version.  See the file
+%   COPYING included with this distribution for more information.
+%   
+%%
+
+
+
+%   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.
+%   Here we are loading test_image.mat that contains structure with 5 images : lena,
+%   barbara,boat, house and peppers.
+clear;
+TMPpath=pwd;
+FS=filesep;
+[pathstr1, name, ext, versn] = fileparts(which('SMALLboxSetup.m'));
+cd([pathstr1,FS,'data',FS,'images']);
+load('test_image.mat');
+cd(TMPpath);
+
+%   Deffining the noise levels that we want to test
+
+noise_level=[10 20 25 50 100];
+
+%   Here we loop through different noise levels and images 
+
+for noise_ind=1:1
+for im_num=1:1
+
+% Defining Image Denoising Problem as Dictionary Learning
+% Problem. As an input we set the number of training patches.
+
+SMALL.Problem = generateImageDenoiseProblem(test_image(im_num).i, 40000, '',256, noise_level(noise_ind));
+SMALL.Problem.name=int2str(im_num);
+
+Edata=sqrt(prod(SMALL.Problem.blocksize)) * SMALL.Problem.sigma * SMALL.Problem.gain;
+maxatoms = floor(prod(SMALL.Problem.blocksize)/2);
+
+%   results structure is to store all results
+
+results(noise_ind,im_num).noisy_psnr=SMALL.Problem.noisy_psnr;
+
+%%
+%   Use KSVD Dictionary Learning Algorithm to Learn overcomplete dictionary
+%   Ron Rubinstein implementation
+
+%   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.
+
+
+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 - find the sparse solution in the learned
+%   dictionary for all patches in the image and the end it uses
+%   reconstruction function to reconstruct the patches and put them into a
+%   denoised image
+
+SMALL.solver(1)=SMALL_solve(SMALL.Problem, SMALL.solver(1));
+
+%   Show PSNR after reconstruction
+
+SMALL.solver(1).reconstructed.psnr
+
+%%
+%   Use KSVD Dictionary Learning Algorithm to Learn overcomplete dictionary
+%   Boris Mailhe ksvd update implentation omp is the same as with Rubinstein
+%   implementation
+
+
+%   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); 
+
+%   Initialising Dictionary structure
+%   Setting Dictionary structure fields (toolbox, name, param, D and time)
+%   to zero values
+
+SMALL.DL(2)=SMALL_init_DL('TwoStepDL', 'KSVD', '', 1);
+
+
+%   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.
+
+
+SMALL.DL(2).param=struct(...
+    'solver', SMALL.solver(2),...
+    'initdict', SMALL.Problem.initdict,...
+    'dictsize', SMALL.Problem.p,...
+    'iternum', 20,...
+    'show_dict', 1);
+
+%   Learn the dictionary
+
+SMALL.DL(2) = SMALL_learn(SMALL.Problem, SMALL.DL(2));
+
+%   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(2).D;
+SMALL.Problem.reconstruct = @(x) ImgDenoise_reconstruct(x, SMALL.Problem);
+
+%   Denoising the image - find the sparse solution in the learned
+%   dictionary for all patches in the image and the end it uses
+%   reconstruction function to reconstruct the patches and put them into a
+%   denoised image
+
+SMALL.solver(2)=SMALL_solve(SMALL.Problem, SMALL.solver(2));
+
+
+%% show results %%
+
+SMALL_ImgDeNoiseResult(SMALL);
+
+clear SMALL;
+end
+end
+
diff -r 485747bf39e0 -r af307f247ac7 examples/Image Denoising/SMALL_ImgDenoise_DL_test_TwoStep_KSVD_MOD_OLS_Mailhe.m
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/examples/Image Denoising/SMALL_ImgDenoise_DL_test_TwoStep_KSVD_MOD_OLS_Mailhe.m	Fri Jul 29 12:35:52 2011 +0100
@@ -0,0 +1,309 @@
+%%  Dictionary Learning for Image Denoising - KSVD vs Recursive Least Squares
+%
+%   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.
+%   Two 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.
+%   -   RLS-DLA - Skretting, K.; Engan, K.; , "Recursive Least Squares
+%       Dictionary Learning Algorithm," Signal Processing, IEEE Transactions on,
+%       vol.58, no.4, pp.2121-2130, April 2010
+%
+
+
+%   Centre for Digital Music, Queen Mary, University of London.
+%   This file copyright 2011 Ivan Damnjanovic.
+%
+%   This program is free software; you can redistribute it and/or
+%   modify it under the terms of the GNU General Public License as
+%   published by the Free Software Foundation; either version 2 of the
+%   License, or (at your option) any later version.  See the file
+%   COPYING included with this distribution for more information.
+%   
+%%
+
+
+
+%   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.
+%   Here we are loading test_image.mat that contains structure with 5 images : lena,
+%   barbara,boat, house and peppers.
+clear;
+TMPpath=pwd;
+FS=filesep;
+[pathstr1, name, ext, versn] = fileparts(which('SMALLboxSetup.m'));
+cd([pathstr1,FS,'data',FS,'images']);
+load('test_image.mat');
+cd(TMPpath);
+
+%   Deffining the noise levels that we want to test
+
+noise_level=[10 20 25 50 100];
+
+%   Here we loop through different noise levels and images 
+
+for noise_ind=2:2
+for im_num=2:2
+
+% Defining Image Denoising Problem as Dictionary Learning
+% Problem. As an input we set the number of training patches.
+
+SMALL.Problem = generateImageDenoiseProblem(test_image(im_num).i, 40000, '',256, noise_level(noise_ind));
+SMALL.Problem.name=int2str(im_num);
+
+Edata=sqrt(prod(SMALL.Problem.blocksize)) * SMALL.Problem.sigma * SMALL.Problem.gain;
+maxatoms = floor(prod(SMALL.Problem.blocksize)/2);
+
+
+%%
+%   Use KSVD Dictionary Learning Algorithm to Learn overcomplete dictionary
+%   Boris Mailhe ksvd update implentation omp is the same as with Rubinstein
+%   implementation
+
+
+%   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); 
+
+%   Initialising Dictionary structure
+%   Setting Dictionary structure fields (toolbox, name, param, D and time)
+%   to zero values
+
+SMALL.DL(1)=SMALL_init_DL('TwoStepDL', 'KSVD', '', 1);
+
+
+%   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.
+
+
+SMALL.DL(1).param=struct(...
+    'solver', SMALL.solver(1),...
+    'initdict', SMALL.Problem.initdict,...
+    'dictsize', SMALL.Problem.p,...
+    'iternum', 20,...
+    'show_dict', 1);
+
+%   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);
+
+%   Denoising the image - find the sparse solution in the learned
+%   dictionary for all patches in the image and the end it uses
+%   reconstruction function to reconstruct the patches and put them into a
+%   denoised image
+
+SMALL.solver(1)=SMALL_solve(SMALL.Problem, SMALL.solver(1));
+
+%%
+%   Use MOD Dictionary Learning Algorithm to Learn overcomplete dictionary
+%   Boris Mailhe MOD update implentation omp is the Ron Rubinstein
+%   implementation
+
+
+%   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); 
+
+%   Initialising Dictionary structure
+%   Setting Dictionary structure fields (toolbox, name, param, D and time)
+%   to zero values
+
+SMALL.DL(2)=SMALL_init_DL('TwoStepDL', 'MOD', '', 1);
+
+
+%   Defining the parameters for MOD
+%   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
+
+SMALL.DL(2).param=struct(...
+    'solver', SMALL.solver(2),...
+    'initdict', SMALL.Problem.initdict,...
+    'dictsize', SMALL.Problem.p,...
+    'iternum', 20,...
+    'show_dict', 1);
+
+%   Learn the dictionary
+
+SMALL.DL(2) = SMALL_learn(SMALL.Problem, SMALL.DL(2));
+
+%   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(2).D;
+SMALL.Problem.reconstruct = @(x) ImgDenoise_reconstruct(x, SMALL.Problem);
+
+%   Denoising the image - find the sparse solution in the learned
+%   dictionary for all patches in the image and the end it uses
+%   reconstruction function to reconstruct the patches and put them into a
+%   denoised image
+
+SMALL.solver(2)=SMALL_solve(SMALL.Problem, SMALL.solver(2));
+%%
+%   Use OLS Dictionary Learning Algorithm to Learn overcomplete dictionary
+%   Boris Mailhe ksvd update implentation omp is the Ron Rubinstein
+%   implementation
+
+
+%   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); 
+
+%   Initialising Dictionary structure
+%   Setting Dictionary structure fields (toolbox, name, param, D and time)
+%   to zero values
+
+SMALL.DL(3)=SMALL_init_DL('TwoStepDL', 'ols', '', 1);
+
+
+%   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.
+
+
+SMALL.DL(3).param=struct(...
+    'solver', SMALL.solver(3),...
+    'initdict', SMALL.Problem.initdict,...
+    'dictsize', SMALL.Problem.p,...
+    'iternum', 20,...
+    'learningRate', 0.1,...
+    'show_dict', 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)
+
+SMALL.Problem.A = SMALL.DL(3).D;
+SMALL.Problem.reconstruct = @(x) ImgDenoise_reconstruct(x, SMALL.Problem);
+
+%   Denoising the image - find the sparse solution in the learned
+%   dictionary for all patches in the image and the end it uses
+%   reconstruction function to reconstruct the patches and put them into a
+%   denoised image
+
+SMALL.solver(3)=SMALL_solve(SMALL.Problem, SMALL.solver(3));
+%%
+%   Use Mailhe Dictionary Learning Algorithm to Learn overcomplete dictionary
+%   Boris Mailhe ksvd update implentation omp is the Ron Rubinstein
+%   implementation
+
+
+%   Initialising solver structure
+%   Setting solver structure fields (toolbox, name, param, solution,
+%   reconstructed and time) to zero values
+
+SMALL.solver(4)=SMALL_init_solver;
+
+% Defining the parameters needed for image denoising
+
+SMALL.solver(4).toolbox='ompbox';
+SMALL.solver(4).name='omp2';
+SMALL.solver(4).param=struct(...
+    'epsilon',Edata,...
+    'maxatoms', maxatoms); 
+
+%   Initialising Dictionary structure
+%   Setting Dictionary structure fields (toolbox, name, param, D and time)
+%   to zero values
+
+SMALL.DL(4)=SMALL_init_DL('TwoStepDL', 'mailhe', '', 1);
+
+
+%   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.
+
+
+SMALL.DL(4).param=struct(...
+    'solver', SMALL.solver(4),...
+    'initdict', SMALL.Problem.initdict,...
+    'dictsize', SMALL.Problem.p,...
+    'iternum', 20,...
+    'learningRate', 2,...
+    'show_dict', 1);
+
+%   Learn the dictionary
+
+SMALL.DL(4) = SMALL_learn(SMALL.Problem, SMALL.DL(4));
+
+%   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(4).D;
+SMALL.Problem.reconstruct = @(x) ImgDenoise_reconstruct(x, SMALL.Problem);
+
+%   Denoising the image - find the sparse solution in the learned
+%   dictionary for all patches in the image and the end it uses
+%   reconstruction function to reconstruct the patches and put them into a
+%   denoised image
+
+SMALL.solver(4)=SMALL_solve(SMALL.Problem, SMALL.solver(4));
+
+%% show results %%
+
+SMALL_ImgDeNoiseResult(SMALL);
+
+%clear SMALL;
+end
+end
+