How do we upsample a colour image by 0.25 and 0.5 without using imresize() function ???

I wanted to upsample a colour (rgb)image by 0.25 and 0.5 without using imresize(). I tried doing it but the output is black & white. Is there any solution ???

This is my script :

%%Lena image Upsample

% Importing original Lena Colour Image

lena_q = imread('lena_quarter.tif');

% Display original Lena image

figure(4)

imshow(lena_q)

title('Original Lena Colour Image')

% Upsample the original Lena image by quarter

scale3=0.25;

lena_quarter=scale3;

lena_quarterup=lena_q(1:lena_quarter:height,1:lena_quarter:width);

figure(5)

imshow(lena_quarterup)

title('Upsizing Original image of the Lena by a quarter')

% Upsample the original Lena image by half

scale4=0.5;

lena_half=scale4;

lena_halfup=lena_q(1:lena_half:height,1:lena_half:width);

figure(6)

imshow(lena_halfup)

title('Upsizing Original image of the Lena by a half')

Thanks

Answer 1

[Walter Roberson]

lena_quarterup = lena_q(1:lena_quarter:height, 1:lena_quarter:width, :);

[ejb the beast]

this gives an error

Answer 2

[Samir Mitha]

You may want to try using the Gridded Interpolant function

Answer 3

[DGM]

Direct subscript indexing isn't going to work, since there are no fractional indices in an array.

lena_quarterup = lena_q(1:lena_quarter:height, 1:lena_quarter:width, :);

For simple image resizing without IPT imresize(), you can just interpolate the image just like any other data.

inpict = imread('peppers.png');
inpict = im2double(inpict);

scale = 2; % upscaling
method = 'bilinear';

% input and output size
[s1 s2 s3] = size(inpict);
s = [s1 s2 s3]; % imsize() avoids this nonsense
os0 = s(1:2)*scale;

% input coordinate space
x0 = linspace(1,s(2),s(2));
y0 = linspace(1,s(1),s(1));
[X0 Y0] = meshgrid(x0,y0);

% output coordinate space
x = linspace(1,s(2),os0(2));
y = linspace(1,s(1),os0(1));
[XX YY] = meshgrid(x,y);

% interpolate
outpict = zeros([os0 s(3)]);
for c = 1:s(3)
    outpict(:,:,c) = interp2(X0,Y0,inpict(:,:,c),XX,YY,method);
end

[size(inpict); size(outpict)] % input/output geometry corresponds to scale factor

ans = 2×3

384 512 3 768 1024 3

imshow(outpict)

Bear in mind that this isn't how imresize() does the task, and the results will not be identical, especially if the scale is <1. For scale <1, imresize() performs antialiasing during interpolation. No such antialiasing is implemented here.

The value of this example depends on whether you want to understand some basic interpolation tasks (e.g. for homework), or if you simply need to resize images without Image Processing Toolbox. If the latter describes your needs, consider that MIMT has imresizeFB(), which can serve as a drop-in replacement for imresize(). It still uses a different interpolation approach than imresize(), but some effort has been made to make the features and results as close as possible.