How to save patches of all images in a single .mat file?

i have a code that divides image into patches of size 32*32 and also divide the corresponding labeled image into same size of patches. i want to save and concatenate all the patches of all the images in a .mat file, but my code is saving only the patches of last image. please can someone help me with this problem?

a = dir('C:\Users\admin\Desktop\img\*.png');
b = dir('C:\Users\admin\Desktop\lbl\*.png');
for i = 1:length(a)
    filename=strcat('C:\Users\admin\Desktop\img\',a(i).name);
    image=imread(filename);
    R = image(:,:,1);
    G = image(:,:,2);
    B = image(:,:,3);
    Y = 0.299 * R + 0.587 * G + 0.114 * B;
    U = -0.14713 * R - 0.28886 * G + 0.436 * B;
    V = 0.615 * R - 0.51499 * G - 0.10001 * B;
    image = cat(3,Y,U,V);
    [image_patches] = hellopatches(image, 32, 20);
    filenamel=strcat('C:\Users\admin\Desktop\lbl\',b(i).name);
    label=imread(filenamel);
    [labels] = hellopatches_label(label, 32, 20);
    data = cat(1, image_patches,labels);
end
data = data';
 data = shuffleRow(data);
save data.mat;

Answer 1

[neuroDuck]

Hi Hirra,

You are overwriting 'data' every time the loop runs. Simply index it - e.g. data(i) - or add it to a cat() to make every image accessible outside of the loop.

Kind regards.

[hirra awan]

But if i index them like this save(['E:\Thesis\3.thesis experiments\Data' num2str(i) '.mat'],'data') every image data would be saved in multiple data files.. i.e data1.mat, data2.mat .. but i want to save all the data in a single data.mat file. how this is possible? please if you can help?

[neuroDuck]

I meant that you would index within the loop, i.e. something like data(i) = cat(1, image_patches,labels); (assuming the data were all the same lengths, otherwise pad with NaNs or save in a cell array, e.g. data{i})

Then you would save all your data after the loop (as you are doing, but the data variable would then contain all your images). Hope this helps. Please accept answer if so, or ask more questions if not.

[hirra awan]

if i do data{i} then the file contains the patches number, to view all the patches of single image i have to click on that. i attached a photo please see.

[Walter Roberson]

if i do data{i} then the file contains the patches number, to view all the patches of single image i have to click on that

What is the problem with that? If we were to save them as individual variables in the .mat file, you would have to click on the individual variable.

[neuroDuck]

Exactly as Walter suggests, what would be the issue with this? Surely the point of wanting to save them all is that you have all the images in a single file that you can access later? This enables you to do exactly that. If you wanted to access them programatically later on, you would just specify the index number of the image you wanted.

[hirra awan]

no i dont want to save my data like this, because from data.mat file i want to train my system of image classification.

[neuroDuck]

What format does your image classifier want? Surely this would guide how you store the data. For example, if I was training an SVM (which I doubt you will for your task, but it's an example), I would want my data to be in samples x features, so I would arrange my data this way when saving it. The way you choose to store your data should be guided by the format of the input you want to put into your image classification.

[hirra awan]

I want my data pixel wise. i am using deep belief network.

[Walter Roberson]

At present, you create a 3073 x 5625 array for the first image, and a 3073 x 5184 array for the second image.

We know from the above discussion that you do not want to put those two arrays as seperate entries in a cell array.

How would you like to store the 3073 x 5625 array together with the 3073 x 5184 arrary ? Would you want the result to be a 3073 x 10809 array?

I cannot seem to find your source code for hellopatches in the discussion, so I have attached it. It appears to me that since your images are not all the same size, that we should not be necessarily be expecting that the output of hellopatches will always have 3073 rows or 5625 columns.

[hirra awan]

Yes exactly I want to store 3073*5625 with 3073*5184 array. This is the data of just 2 images. i have almost 160 images in total. I want the final .mat file to store data like this 3073*10809 .

I attached the hello patches code please check.

Answer 2

[Walter Roberson]

    data = cat(1, image_patches,labels);

That code is overwriting all of data each iteration.

It looks to me as if data will not be a vector, but I do not seem to be able to locate any hellopatches() function so I cannot tell what shape it will be. As you are not doing imresize() I also cannot be sure that all of the images are the same size, so I cannot be sure that data will be the same size for each iteration. Under the circumstances you should be considering saving into a cell array.

Note: please do not post the same query multiple times. I found at least 9 copies of your query :(

[Walter Roberson]

We need one sample image from each of the two directories to test with.

[hirra awan]

these are the 2 images. one is tissue image and other is the corresponding label image.

[hirra awan]

this is the labeled image.

[Walter Roberson]

We also need the source code for hellopatches_label

[hirra awan]

it is same as the hellopatches code

[hirra awan]

if i have multiple .mat files with pixel values of images, how can i copy all my data into single .mat file? please let me know if you have answer?

i want to save my data like this as shown in the image.. each .mat fil contains this type of data.

[Walter Roberson]

if ispc()
    imgdir = 'C:\Users\admin\Desktop\img';
    labeldir = 'C:\Users\admin\Desktop\lbl';
else
    imgdir = 'img';
    labeldir = 'lbl';
end

a = dir( fullfile(imgdir, '*.png') );
b = dir( fullfile(labeldir, '*.png') );

data = [];
for i = 1:length(a)
    filename = fullfile(imgdir, a(i).name);
    RGB = imread(filename);
    R = RGB(:,:,1);
    G = RGB(:,:,2);
    B = RGB(:,:,3);
    Y = 0.299 * R + 0.587 * G + 0.114 * B;
    U = -0.14713 * R - 0.28886 * G + 0.436 * B;
    V = 0.615 * R - 0.51499 * G - 0.10001 * B;
    YUV = cat(3,Y,U,V);
    [image_patches] = hellopatches(YUV, 32, 20);
    filenamel = fullfile(labeldir, b(i).name);
    label = imread(filenamel);
    [labels] = hellopatches_label(label, 32, 20);
    data = cat(1, data, image_patches,labels);
end
data = data';
 data = shuffleRow(data);
save data.mat;

[hirra awan]

data = cat(1, data, image_patches,labels);

this line is not correct its giving error.

data = cat(1,image_patches,labels);// here i have writtten this line because i want to concatinate the pixel values of origional image patches and the corresponding labeled image patches in form of columns.

but you have added data in it so its not working..issue remains the same :(

[Walter Roberson]

// is not a permitted comment marker in MATLAB.

[hirra awan]

i know.. // these are not in the code, i put them here to elaborate my code and problem.