How to divide images from folder into 4x4 blocks

I want to divide all images in a folder into 4x4 blocks and then store the RGB information in a zeros matrix. A sample of what the images look like is attached. Does anyone know how to divide into these blocks and then store the information in the matrix.

[Turlough Hughes]

"and then store the RGB information in a zeros matrix", can you expand on that. How exactly do the 4x4 pixel blocks relate to your new matrix?

[Turlough Hughes]

Also, consider the blockproc function. This might be the way to go depending on what it is you're trying to do.

[Petrus van Aswegen]

Sorry for the poor wording mate. The end state is to use the RGB colour information in each 4x4 patch to isolate the properties of the solar panels, and then use support vector machine to detect solar panels in other images. I basically need to divide the images into patches, and then store the information within each patch into a matrix

Answer 1

[Turlough Hughes]

Hi Petrus, the mat2cell approach that Jonas pointed out can be done more generally as follows. Note, this handles the scenario where image dimensions are not a factor of the corresponding block dimensions (the remainder rows or columns are ommited)

1. Parameters

blockHeight = 4;
blockWidth = 4;
folderName = 'C://...'; % folder containing images

2. Directory. These three lines can be used by others to replicate the demo. Once you modify folderName you can replace the following three lines with the fourth one.

[fils(1:6).folder] = deal('');
fnames = cellfun(@(x) sprintf('office_%s.jpg',x),{'1','2','3','4','5','6'},'uni',false);
[fils.name] = fnames{:};
%fils = dir(fullfile(folderName,'*.PNG'));  < Replace the above with this line

3. Load an initial image to get some metadata:

I0 = imread(fullfile(fils(1).folder,fils(1).name));
N1 = blockHeight*ones(floor(size(I0,1)/blockHeight),1);
N2 = blockWidth*ones(floor(size(I0,2)/blockWidth),1);

4. Loop through your directory:

L = cell(numel(fils),1);
for ii = 1:numel(fils)
    I = imread(fullfile(fils(ii).folder,fils(ii).name));
    b = mat2cell(I(1:blockHeight*numel(N1),1:blockWidth*numel(N2),:),N1,N2,3);
    L{ii} = cat(4,b{:}); % alternatively L{ii} = b;
end
L

L = 6×1 cell array
    {4×4×3×33750 uint8}
    {4×4×3×33750 uint8}
    {4×4×3×33750 uint8}
    {4×4×3×33750 uint8}
    {4×4×3×33750 uint8}
    {4×4×3×33750 uint8}

The result, L, is then a 6 by 1 cell array of matrices where each cell corresponds to an image. The matrix dimensions (4x4x3x3750) correspond to the block height, block width, RGB, and a linear index for the block.

[Petrus van Aswegen]

This worked perfectly, thank you so much mate. Appreciate your help

Answer 2

[Jonas]

you could use mat2cell

A = rand(256,256,3); % your matrix here
N = 4*ones(1,64);
B = mat2cell(A,N,N,3);

adjusted from here

https://de.mathworks.com/matlabcentral/answers/89757-how-to-divide-256x256-matrix-into-sixteen-16x16-blocks

[Petrus van Aswegen]

Thanks mate. Do you know how i could apply this to all of the images in a folder. I have 192 images i want divide into patches and i assume some "for" loop would be best but not sure how to formulate this

[Jonas]

catch all files you are interested in eg by {dir('*.jpg').name} and loop through the file names and open them with imread()

[Petrus van Aswegen]

Legend. Thanks for the help mate, much appreciated.