CODEWITHSUNDEEP
matlab
Michael Michael Michael
Michael Michael Michael

Reputation: 159

Converting an image (3D array) into a 2D matrix in a specific order

Say you have an image with 200 pixels. It will be a 10 x 20 x 3 array where the pages are layer of colors (Red, green and blue). How do I convert that, into a 3 (row) x N (columns eg. 2000) so that each row represents the color (row 1 is red, row 2 is green etc.) and the columns represent pixels

I've tried reshaping but I get a 3 by N matrix that fills the rows downwards and not horizontally (so each row is a mixture of colors and not a specific color).

Upvotes: 0

Views: 422

Answers (1)

HansHirse
HansHirse

Reputation: 18895

Your idea using reshape is correct, but as you found out yourself, the order of the array dimensions is important. Luckily, you can manipulate this using permute. So, in your case, the "color information", i.e. the third dimension, should be set to the first dimension, so that reshape works as intended.

Let's have look at this code snippet:

% Set up dimensions
rows = 10;
cols = 20;

% Generate artificial image
img = uint8(255 * rand(rows, cols, 3));

% Get color independently for each channel
r = reshape(img(:, :, 1), 1, rows * cols);
g = reshape(img(:, :, 2), 1, rows * cols);
b = reshape(img(:, :, 3), 1, rows * cols);

% Reshape image with previous dimension permuting
img2 = reshape(permute(img, [3 1 2]), 3, rows * cols);

% Compare results
rOK = (sum(r == img2(1, :)) == rows * cols)
gOK = (sum(g == img2(2, :)) == rows * cols)
bOK = (sum(b == img2(3, :)) == rows * cols)

For a simple comparison, I fetched the "color information" separately, cf. the vectors r, g, and b. Then I permuted the original img as described above, reshaped it to a 3 x N matrix as desired, and compared each row with r, g, and b.

Hope that helps!

Upvotes: 2

Related Questions