Why does the size of third dimension of a 3D array change?

I have an image with dimension 124x1422x3 which needs to be binarized. When i access it with a simple imread() function and see the dimension and size of the third dimension, it shows 3 in both cases which is true. But when i try to access it with imread() function again but this time to binarize it with the im2bw() function, it throws an error saying "Error using im2bw>parse_inputs (line 97) IM2BW: Truecolor RGB image has to be an M-by-N-by-3 array." When i checked the dimension of the array it shows 124x1422x4 and i am clueless as to why the size of 3rd dimension is changing when all i am doing is the same except for sending the image for binarization.Any help is appreciated. Thanks in advance.

[James Tursa]

Please show us your code.

[deena hijam]

for k = 2 : numberOfFolders
  % Get this folder and print it out.

      l=1;
    thisFolder = listOfFolderNames{k};
    fprintf('Processing folder %s\n', thisFolder);

     folder = [Datapath,sprintf('%s_cropped',thisFolder(25:end))];
      %if not(exist(folder_name,'dir'))
          mkdir(folder);
    
    % Get PNG files.
  filePattern = sprintf('%s/*.png', thisFolder);
    baseFileNames = dir(filePattern);
    % Add on TIF files.
    filePattern = sprintf('%s/*.tif', thisFolder);
    baseFileNames = [baseFileNames; dir(filePattern)];
    % Add on JPG files.
  %   filePattern = sprintf('%s/*.jpg', thisFolder);
  %   baseFileNames = [baseFileNames; dir(filePattern)];
    numberOfImageFiles = length(baseFileNames);
    % Now we have a list of all files in this folder.
    
    if numberOfImageFiles >= 1
      % Go through all those image files.
      for f = 1 : numberOfImageFiles
        fullFileName = fullfile(thisFolder, baseFileNames(f).name);
        fprintf('     Processing image file %s\n', fullFileName);
          end

   for j=1:numberOfImageFiles
      st = sprintf('%04d.tif', j);
      %data(j).im =imread(fullfile(thisFolder, st));
      data =imread(fullfile(thisFolder, st));

dimension1 = ndims(data);
sizeof1 = size(data,3);

   %Thresholding using Otsu's method
  threshold = graythresh(data);
  bw = im2bw(data,threshold);
  %imshow(BW);
  %hold on
  %figure(2)
  bw1=~bw;

   se = strel('square', 11);
  im_close = imclose(bw1,se);

imlabel= bwlabel(im_close);
%CC = bwconncomp(bw1,26)
s = regionprops(imlabel, 'BoundingBox');
bb = round(reshape([s.BoundingBox], 4 , []).');
[L Ne]=bwlabel(im_close);

%Ne = CC.NumObjects; 
%imshow(BW1);
for idx = 1:numel(s)
   % rectangle('Position', bb(idx,:), 'edgecolor', 'red');
end

blobMeasurements = regionprops(imlabel, 'BoundingBox');
 for m=1:Ne
BB(m,:) = blobMeasurements(m).BoundingBox;
 end
 %txt=['A' 'B' 'C' 'D' 'E' 'F' 'G' 'H' 'I' 'J' 'K' 'L' 'M' 'N' 'O' 'P' 'Q' 'R' 'S' 'T' 'U' 'V' 'W' 'X' 'Y' 'Z' '1' '2' '3' '4']; 
% txt is Random matrix of character for naming the files for saving them

%se1 = strel('ball',5,5); %for dilation
 for i= 1:Ne
imagen{i} = imcrop(data, [BB(i,1) BB(i,2) BB(i,3) BB(i,4)]);

paddedImage{i} = padarray(imagen{i},[10 10],255,'both');

   %mkdir(sprintf('E:\folder%d',k)); 
   %folder=[Datapath1,num2str(k)]
    %folder = 'E:\folder';
   baseFileName = sprintf('%04d.tif', l);
  fullFileName = fullfile(folder, baseFileName);
  %imwrite(resizedImage{i}, fullFileName);
  imwrite(paddedImage{i}, fullFileName,'TIFF','Resolution',[300 300]) %Xresolution and Yresolution of 300dpi
  l=l+1;

   end
   %k=k+l-2;
   end
      end
  end

I have a folder containing a number of subfolders. What i am trying to do is to go to each subfolder and access each image file in the subfolder and crop out the characters from those image files. And then save the cropped character images in another folder.

[Walter Roberson]

To check, then: the difficulty is that the .tif file that you create in the *_cropped folder has somehow become something-by-something-by-4 even though the data that you are processing to create that .tif is -by-3 ?

[deena hijam]

I am trying to crop the characters from the image files inside the subfolders. In order to do that, i am binarizing the images first. But since somehow the .tif has changed from something-by-something-by-3 to something-by-something-by-4, the im2bw() function is showing the error "Error using im2bw>parse_inputs (line 97) IM2BW: Truecolor RGB image has to be an M-by-N-by-3 array." There are a total of 253 .tif files to be cropped out of which the first 26 are fine. The 27th one is the problematic one with dimension -by-4. The .tif files in the original folder are a mix of something-by-something-3 uint8 and something-by-something uint8. This is also the same for the .tif files cropped out from the first 26 original .tif files in the *_cropped folders. If i could fix the dimension problem of the 27th file and other such files which i am sure there must be, my problem is solved for now. Thanks.

[deena hijam]

I think the problem is with the bit depth of the images. The images that are causing problem have a bit depth of 32 bits. Is there any way i can change the bit depth from 32 to say 24 bits for TIFF files?? I am using MATLAB R2015a version.

[Walter Roberson]

Your description talks about there being two imread() of the same file, and talks about im2bw() being the second of them. But your posted code contains only

        data =imread(fullfile(thisFolder, st));

dimension1 = ndims(data);
sizeof1 = size(data,3);

     %Thresholding using Otsu's method
    threshold = graythresh(data);
    bw = im2bw(data,threshold);

which is a single imread() followed by im2bw()

This leaves me confused about where the "first" imread() is that is successful in identifying it as -by-3 .

Would it be possible for you to attach one of the images that is causing the problem?

[deena hijam]

Hi Walter, I am sorry for the late response and also for not mentioning that the other imread() is a different piece of code that i have written just to check what it reads. Also the image file was converted to .jpg file before being read which i don't know why i did. But now when i try to read the same file once by the original .tif and other with the .jpg format, the dimensions are the same but the sizes are different. Does it have to do anything with the format which i save it with. I am totally confused now. Below I am attaching a screenshot of the code and the output:

And the .tif file causing the problem is also attached. Since the .tif format is not supported, i could not attach it.

[Walter Roberson]

If the .tif is the original and you are converting it to jpg and the tif is x4 and the jpg is x3, then that would suggest that the issue is with the original tif. I would need to have a look at the original tif.

[deena hijam]

Since .tif format is not supported, i am sending a .zip file

Answer 1

[Walter Roberson]

Okay, so there are two ways you can handle these:

1) make some assumptions that are probably valid:

im0027 = imread('0027.tif');
if size(im0027,3) == 4; im0027 = im0027(:,:,3); end

2) don't make assumptions

First download http://www.tech-view.com/download/ICC_profiles/USSheetfedCoated.icc and make sure it is on your MATLAB path. Or you might be able to get it from https://www.mathworks.com/matlabcentral/answers/101194-how-can-i-convert-cmyk-image-to-rgb#answer_110542

The MATLAB documentation tends to imply that USSheetfedCoated.icc should be available without that, but if you read it carefully really you have to download it, and the Mathworks documentation gives the wrong location for it.

Then:

filename = '0027.tif';
info = imfinfo(filename);

if isfield(info, 'PhotometricInterpretation') && strcmpi(info.PhotometricInterpretation, 'CMYK')
  outprof = iccread('sRGB.icm');
  inprof = iccread('USSheetfedCoated.icc');
  C = makecform('icc',inprof,outprof);
  im0027 = applycform(imread(filename), C);   %CMYK input
  im0027_alpha = [];
elseif isfield(info, 'ExtraSamples') && info.ExtraSamples ~= 0   %1 -- associated alpha; 2 -- unassociated alpha
  [im0027, im0027_alpha] = readRGBAImage(Tiff(filename, 'r'));
elseif isfield(info, 'BitsPerSample')
  chans = length(info.BitsPerSample);
  if chans == 1
    [im0027, im0027_map] = imread(filename);     %grayscale or pseudocolor
    if ~isempty(im0027_map); im0027 = ind2rgb(im0027, im0027_map); end   %pseudocolor
    im0027_alpha = [];
  elseif chans == 2
     error('file "%s" has two channels, which is probably a broken implementation of a binary image with an alpha channel. Giving up on it.', filename);
  elseif chans == 3
    im0027 = imread(filename);     %rgb
    im0027_alpha = [];
  elseif chans == 4
    warning('file "%s" has four channels but is not CMYK or correct RGBA, using first three', filename);
    im0027 = imread(filename);     %rgb+something
    im0027 = im0027(:,:,1:3);
  elseif chans == 5
    warning('file "%s" has five channels but is not CMYK or correct RGBA, using first three even though it is probably hyperspectral', filename);
    im0027 = imread(filename);     %rgb+something+something
    im0027 = im0027(:,:,1:3);
  else
    error('file "%s" unexpectedly has %d channels, must be hyperspectral, giving up on it', filename, chans);
  end
else         %hmmm, what _is_ it anyhow?
  [im0027, im0027_map] = imread(filename);     %unmarked RGBm, or grayscale or pseudocolor
  if ~isempty(im0027_map); im0027 = ind2rgb(im0027, im0027_map); end   %pseudocolor
  im0027_alpha = [];
  if ndims(im0027) > 3 || ~ismember(size(im0027,3), [1, 3])
    error('file "%s" does not have expected headers describing it, and is not grayscale or RGB, giving up on it', filename);
  end
end

The above covers RGB, RGBA with proper headers, CMYK, RGBA without proper headers (under protest), grayscale, pseudocolor, RGB plus two hyperspectral channels (under protest)

[deena hijam]

That worked! Thanks a ton for the help Walter. Now all the files with x4 are with x3. One more thing, if the original images are grayscale is there a way to convert it to RGB. I want all the files to be normalized to x3 format.

[deena hijam]

I think i got the solution to this as well from one of the answers you provided. Thank you so much again. :)

[Walter Roberson]

The easiest way to convert grayscale to RGB is to repmat(TheArray, [1 1 3])

[deena hijam]

Yes. Thank you