Alter and store images back into an excisting imagedatastore

Hi All,

I'm new to Matlab and need some help on how to alter and store images back into an (excisting) imagedatastore.

I already figured out how to read a bunch of images (which are labeled) and then, based on the largest object in the image, calculate and crop/resize the image for further processing.

I now need to update the original image in the imagedatastore with the new crop/resized image. Is this possible and how.

Kind regards

Bert

Code :

ORIGINAL = imageDatastore(digitDatasetPath,...
    'IncludeSubfolders',true,'LabelSource','foldernames');

while hasdata(ORIGINAL)

% Clean up image
B = read(ORIGINAL);
BW = rgb2gray(B);
BW2 = imbinarize(BW);

% Label the image
labeledImage = bwlabel(BW2);
measurements = regionprops(BW2, 'BoundingBox', 'Area');

% Extract the biggest blob
allAreas = [measurements.Area];
[sortedAreas, sortingIndexes] = sort(allAreas, 'descend');

% Draw boundingbox arround biggest blob
Bigblob = measurements(sortingIndexes(1)).BoundingBox;

% Cut out biggest blob
DLim = imcrop(B,[Bigblob(1) Bigblob(2) Bigblob(3) Bigblob(4)]);

% Square image before resizing to maintain aspect ratio
Wblob = Bigblob(3);
Lblob = Bigblob(4);

if Wblob == Lblob
    % do nothing
else
    if Wblob > Lblob 
        DLim = padarray(DLim,[Wblob-Lblob 0],'pre');
    else
        DLim = padarray(DLim,[0 Lblob-Wblob],'pre');
    end
end

% Resize - this result needs to replace the original image in the ORIGINAL imagedatastore ...
DLim = imresize(DLim,[1024,1024]);

end

Answer 1

[Arun Mathamkode]

I can see that you are reading one image at a time (readSize=1). Hence the easy way is to write the image to the location specified by the imageDatastore object.

Get the file info as the second argument while using the 'read' function. Then use the 'Filename' property in the file info to update the cropped/resized image back to the original location.

      ORIGINAL = imageDatastore(digitDatasetPath,'IncludeSubfolders',true,'LabelSource','foldernames');

      while hasdata(ORIGINAL)
      % Clean up image
      [B,info] = read(ORIGINAL);   % get the File info
      BW = rgb2gray(B);
      BW2 = imbinarize(BW);
      % Label the image
      labeledImage = bwlabel(BW2);
      measurements = regionprops(BW2, 'BoundingBox', 'Area');
      % Extract the biggest blob
      allAreas = [measurements.Area];
      [sortedAreas, sortingIndexes] = sort(allAreas, 'descend');
      % Draw boundingbox arround biggest blob
      Bigblob = measurements(sortingIndexes(1)).BoundingBox;
      % Cut out biggest blob
      DLim = imcrop(B,[Bigblob(1) Bigblob(2) Bigblob(3) Bigblob(4)]);
      % Square image before resizing to maintain aspect ratio
      Wblob = Bigblob(3);
      Lblob = Bigblob(4);
      if Wblob == Lblob
          % do nothing
      else
          if Wblob > Lblob 
              DLim = padarray(DLim,[Wblob-Lblob 0],'pre');
          else
              DLim = padarray(DLim,[0 Lblob-Wblob],'pre');
          end
      end
      % Resize - this result needs to replace the original image in the ORIGINAL imagedatastore ...
      DLim = imresize(DLim,[1024,1024]);
      imwrite(DLim,info.Filename);   % Update the image in the original location

      end

When you are reading multiple image files simultaneously, you may need to use a loop while writing the image.

[Aarti Bokade]

Sir can you please write how to do it for all the images stored in a folder and preprocess all the images and store them back to same folder after preprocessing

Answer 2

[prasanth s]

You can use augmented image datastore method to resize the image before training(matlab 2017b and after).