Contents

Downsize an array using different methods

clear;
% load an RBG image: 135-by-198-by-3 array
I = imread('onion.png');
disp(size(I));
figure; imagesc(I);

   135   198     3

Set target dimension

targetdim = round([100 150 3]);
disp(targetdim);

   100   150     3

Downsize using interpolation

I_interp = array_resize(I, targetdim); %<-- default method is interpolate
disp(size(I_interp));
figure; imagesc(I_interp);

   100   150     3

Downsize using discrete cosine transform (DCT)

I_dct = array_resize(I, targetdim, 'method', 'dct');
disp(size(I_dct));
figure; imagesc(I_dct);

   100   150     3

Downsize to PCA scores using HOSVD. PC scores loose original interpretation

I_hosvd = array_resize(I, targetdim, 'method', 'hosvd');
disp(size(I_hosvd));
figure; imagesc(I_hosvd);

   100   150     3

Downsize to PCA scores using marginal SVD. PC scores loose original interpretation

I_2dsvd = array_resize(I, targetdim, 'method', '2dsvd');
disp(size(I_2dsvd));
figure; imagesc(I_2dsvd);

   100   150     3

Upsize an array using different methods

Set target dimension

targetdim = round([200 300 3]);
disp(targetdim);

   200   300     3

Upsize using interpolation

I_interp = array_resize(I, targetdim); %<-- default method is interpolate
disp(size(I_interp));
figure; imagesc(I_interp);

   200   300     3

Upsize using discrete cosine transform (DCT)

I_dct = array_resize(I, targetdim, 'method', 'dct');
disp(size(I_dct));
figure; imagesc(I_dct);

   200   300     3

Downsizing-analysis-upsizing

clear;
% reset random seed
s = RandStream('mt19937ar','Seed',2);
RandStream.setGlobalStream(s);

True coefficients for regular (non-array) covariates

p0 = 5;
b0 = ones(p0,1);

2D true signal: 64-by-64 cross

shape = imread('cross.gif');
shape = array_resize(shape,[32,32]); % 32-by-32
b = zeros(2*size(shape));
b((size(b,1)/4):(size(b,1)/4)+size(shape,1)-1, ...
    (size(b,2)/4):(size(b,2)/4)+size(shape,2)-1) = shape;
[p1,p2] = size(b);
disp(size(b));

figure; imagesc(-b);
colormap(gray);
title('True Signal');
axis equal;
axis tight;

    64    64

Simulate covariates

n = 500;    % sample size
X = randn(n,p0);   % n-by-p0 regular design matrix
M = tensor(randn(p1,p2,n));  % p1-by-p2-by-n matrix variates
disp(size(M));

    64    64   500

Simulate responses

mu = X*b0 + double(ttt(tensor(b), M, 1:2));
sigma = 1;  % noise level
y = mu + sigma*randn(n,1);

Rank 2 Kruskal regression with 64-by-64 covariates

tic;
[~,beta1,glmstats1] = kruskal_reg(X,M,y,2,'normal');
toc;
disp(size(beta1));

figure;
imagesc(-double(beta1));
colormap(gray);
title('Estimate using 64x64 covariates');
axis equal;
axis tight;

Elapsed time is 1.884143 seconds.
    64    64

Resize 2D covariates to 32-by-32

M_reduce = array_resize(double(M), [32 32 n], 'method', 'dct');
disp(size(M_reduce));

    32    32   500

Rank 2 Kruskal regression with 32-by-32 covariates

tic;
[~,beta2,glmstats2] = kruskal_reg(X,M_reduce,y,2,'normal');
toc;
disp(size(beta2));

Elapsed time is 0.366865 seconds.
    32    32

Resize estimate and disp in original size

beta2 = array_resize(double(beta2), [64 64], 'method', 'dct');
disp(size(beta2));

figure;
imagesc(-double(beta2));
colormap(gray);
title('Estimate using 32x32 covariates');
axis equal;
axis tight;

    64    64

Published with MATLAB® R2017a