4d Array converting RGB image into binary

Hello, I want saved 3d images into a 4d array with following code:

for slice = 1 : length(filestrain)
  filename = fullfile(foldertrain, filestrain(slice).name);
  thisImage = imread(filename);
  [rows, columns, numberOfColorChannels] = size(thisImage);
  if numberOfColorChannels < 3
    message = 'Error: Image is not RGB Color!';
    uiwait(warndlg(message));
    continue;
  end
  if rows ~= 1603 || columns ~= 1603
    message = 'Error: Image is not 1603X1603!';
    uiwait(warndlg(message));
    continue; % Skip this image.
  end
  % Image is okay.  Insert it.
  XTrain(:,:,:,slice) = thisImage;
  imshow(thisImage)
end

The images are zero padded images (1603x1603x3). The original images all have different sizes (cutted manually) and are partly very small like 321x97x3. I did this zero padding for the Input layer of a CNN.

when I am trying to display some images from the 4d Array with:

imshow(XTrain(:,:,2))

The image which is shown is like a binary image no RGB. I dont understand why. Is there a problem with the size of the images?

Thanks for your help

Answer 1

[DGM]

imshow(XTrain(:,:,2))

Is going to be a 4D monochrome image, which imshow won't like. I'm assuming you meant

imshow(XTrain(:,:,:,2))

which is a single frame from the stack.

One likely cause of the thresholding behavior is a mismatch between the image class and its data range. Imshow expects floating point images to be within [0 1]. If they're outside the range, they'll be clipped. This likely happens because you're preallocating your array like so:

XTrain = zeros(1603,1603,3,length(filestrain)); % this is implicitly double

You can pick a class when you preallocate the array (let's say uint8)

XTrain = zeros(1603,1603,3,length(filestrain),'uint8'); % this is explicitly uint8

and then make sure to cast and scale the images as you import them

XTrain(:,:,:,slice) = im2uint8(thisImage);

For what it's worth, MIMT has tools for batch importing into 4D arrays, with automatic padding, casting, and orientation. This loads a mix of monochrome and color images of various sizes and pads them with black to fit.

fprefix = fullfile(matlabroot,'toolbox','images','imdata')
fnames = {'c*.tif'};
instack = mimread(fprefix,fnames,'verbose');
instack = imstacker(instack,'padding',0,'gravity','c', ...
    'outclass','uint8','verbose');

imshow2('instack','tools') % view 4D array

MIMT can be downloaded on the file exchange:

https://www.mathworks.com/matlabcentral/fileexchange/53786-image-manipulation-toolbox

Documentation of mimread/imstacker is here:

http://mimtdocs.rf.gd/manual/html/mimread.html

http://mimtdocs.rf.gd/manual/html/imstacker.html

[forgood2]

Thank you!