Solution to exercise week 7, Univariate Gaussian classifier

Contents

Univariate Gaussian Classifier

A suggested "solution" to classification exercise 1 the univariate Gaussian classifier. This week I will not provide the actual solution, it is hidden in the UnivariateGauss function. Since it is a point that you should implement this yourself. Anyway, I have provided some results so that you have something to compare your code with. If you have large differences from my result, please let me know and we'll check it out.

clear all
close all

The feature images

We are provided with a Landsat sattelite image containing 6 image "bands". We will use each of these bands as a feature image. Lets load them, I'm saving them in a so called cell array.

tm = cell(6);
tm{1} = imread('tm1.png');
tm{2} = imread('tm2.png');
tm{3} = imread('tm3.png');
tm{4} = imread('tm4.png');
tm{5} = imread('tm5.png');
tm{6} = imread('tm6.png');
% We are also going to need these training and test masks. These masks
% indicate what pixels are known to belong to each class. Thus indicating
% what pixels we should use for training, and what pixels we can use to
% validate the result (test).
tm_train = imread('tm_train.png');
tm_test = imread('tm_test.png');

figure(1)
subplot(211)
imshow(tm{1},[]);
colorbar
title('Feature 1');

subplot(212)
imshow(tm{2},[]);
colorbar
title('Feature 2');

figure(2)
subplot(211)
imshow(tm{3},[]);
colorbar
title('Feature 3');

subplot(212)
imshow(tm{4},[]);
colorbar
title('Feature 4');

figure(3)
subplot(211)
imshow(tm{5},[]);
colorbar
title('Feature 5');

subplot(212)
imshow(tm{6},[]);
colorbar
title('Feature 6');

figure(4)
subplot(211)
imagesc(tm_train);
colorbar
axis image
title('Training mask');

subplot(212)
imagesc(tm_test);
colorbar
axis image
title('Test mask');
drawnow

Classification

We are going to use the univariate gaussian classifier, meaning that we are only unsing one feature at a time and classify the image with just one feature. The goal of this exercise is to evaluate each feature and figure out which seems to be the best.

number_of_features = size(tm,2);     %Number of features
k = double(max(tm_train(:)));        %Number of classes
nbr_classified = sum(tm_test(:)>0);  %The total number of pixels in the test set
[N,M] = size(tm{1});                 %Size of image

for feature = 1:number_of_features

    %We are sending the images into this "black box" that you should
    %implement and we are getting the resulting class.
    fig_nbr = 5+feature;
    class = UnivariateGauss(feature,tm_train,tm,fig_nbr);

    %Make a labeled image to display how the pixels were classified
    figure(100+feature);clf
    imagesc(class)
    colormap jet
    colorbar
    title(['Feature: ',num2str(feature),', classification result']);
    drawnow

    %Lest mask out the traning part of the image
    img_labeled_train = zeros(N,M);
    img_labeled_train(tm_train==1) = class(tm_train==1);
    img_labeled_train(tm_train==2) = class(tm_train==2);
    img_labeled_train(tm_train==3) = class(tm_train==3);
    img_labeled_train(tm_train==4) = class(tm_train==4);

    figure(200+feature);
    imagesc(img_labeled_train);
    colormap jet
    colorbar
    title(['Feature: ',num2str(feature),', masked result (train) for the classes']);
    drawnow

    %And the test part of the image
    img_labeled = zeros(N,M);
    img_labeled(tm_test==1) = class(tm_test==1);
    img_labeled(tm_test==2) = class(tm_test==2);
    img_labeled(tm_test==3) = class(tm_test==3);
    img_labeled(tm_test==4) = class(tm_test==4);

    figure(300+feature);
    imagesc(img_labeled)
    colormap jet
    colorbar
    title(['Feature: ',num2str(feature),', masked result (test) for the classes']);
    drawnow

    %Calculate the correct classification
    fprintf('Feature %d: \n',feature);
    correct = zeros(1,k);
    percent = zeros(1,k);
    error = 0;
    %Lets go through each class and calculate
    for class_index = 1:k
        % The error: by summing up how many class labels does not
        % correspond with the class index.
        error = error + sum(class(tm_test==class_index)~=class_index);
        % The correct is thus how many does fit
        correct(class_index) = sum(class(tm_test==class_index)==class_index);
        percent = correct(class_index)/sum(sum(tm_test==class_index));
        fprintf('Class %d correct: %f \n',class_index,percent*100);
    end
    fprintf('Total error: %f\n',error/nbr_classified);
    fprintf('Correct classification for feature image %d is %f \n\n\n',...
        feature,sum(correct)*100/nbr_classified);

end

Feature 1: 
Class 1 correct: 24.060150 
Class 2 correct: 64.305949 
Class 3 correct: 83.223850 
Class 4 correct: 69.657321 
Total error: 0.385892
Correct classification for feature image 1 is 61.410839 


Feature 2: 
Class 1 correct: 26.142279 
Class 2 correct: 95.609065 
Class 3 correct: 91.611925 
Class 4 correct: 60.560748 
Total error: 0.270370
Correct classification for feature image 2 is 72.963008 


Feature 3: 
Class 1 correct: 40.948525 
Class 2 correct: 80.949008 
Class 3 correct: 92.218292 
Class 4 correct: 64.361371 
Total error: 0.280816
Correct classification for feature image 3 is 71.918397 


Feature 4: 
Class 1 correct: 81.954887 
Class 2 correct: 0.000000 
Class 3 correct: 80.545730 
Class 4 correct: 78.380062 
Total error: 0.475359
Correct classification for feature image 4 is 52.464053 


Feature 5: 
Class 1 correct: 47.079237 
Class 2 correct: 66.395184 
Class 3 correct: 96.664982 
Class 4 correct: 70.591900 
Total error: 0.295195
Correct classification for feature image 5 is 70.480521 


Feature 6: 
Class 1 correct: 25.563910 
Class 2 correct: 64.907932 
Class 3 correct: 96.766043 
Class 4 correct: 81.433022 
Total error: 0.324444
Correct classification for feature image 6 is 67.555610 

Published with MATLAB® R2015a