Creating Feature vector in .mat file

I wanted to create a feature vector for training dataset and wanted to store all feature as rows in the.mat file. The .mat file must be in the form

I am able to extract Feature of 1 image and store it in excel file or .mat file but not able to extract all image feature and store it in .mat file. Can anyone knows this?

https://stackoverflow.com/questions/44616804/creating-feature-vector-in-mat-file-for-training-dataset

[KSSV]

Are the dimensions of features of all images same?

[dpb]

Besides KSSV's request for needed info, show us what you've tried...descriptions aren't sufficient to debug.

[dpb]

Don't expect folks to go somewhere else...post the info here or use stackoverflow, your choice.

[Neha singh]

Hi @KSSV, All the dimension of the images are same. So here is my sample code .

I am able to get

xpecting somthing to do like

Answer 1

[dpb]

graycoprops returns a structure containing the requested statistics in fields of each name; hence you got what you requested in the .mat file.

To convert that structure to an array of contiguous features, use

features=struct2array(stats);

if all the features returned are commensurately-sized arrays (and I don't know if that is true or not; I don't have Image Processing TB to test). If not, you'd have to take the variously-shaped/sized arrays in the various structure names and convert them individually to linear vectors to concatenate by row.

In your code above, even when you're concatenating the structs into mydataGLCM, you've got the initialization of creating the empty container variable INSIDE the loop so you end up with only the last one as that throws away all the previous.

PS: Why do you continue to make it so difficult to try to help by pasting essentially nearly illegible images owing to their size plus one can't do anything with the code without typing everything in again. Copy and paste text and code so folks here have a chance (and then be sure to format it so it's legible using the {}Code button where needed.

[Neha Singh]

Thanks for your suggestion @ dpb at that movement I was just having the snap in my mail. I tried this:

    clear;
    file_list = dir('C:\Users\Neha\workspace_matlab\My Project\final 
    project\output\GLCM\*.jpg');
    nfiles = length(file_list);
    mydataGLCM = [];

    for k = 1 : nfiles
        I = imread(strcat('C:\Users\Neha\workspace_matlab\My 
        Project\final project\output\GLCM\',file_list(k).name));

        if length(size(I))== 3 %check if the image I is color
            I=rgb2gray(I);
        end

        GLCM_values = graycomatrix(I);
        stats = graycoprops(GLCM_values{'contrast','homogeneity','Correlation','Energy'});
        Contrast = stats.Contrast;
        Correlation = stats.Correlation;
        Energy = stats.Energy;
        Homogeneity = stats.Homogeneity;
        feat = [Contrast,Correlation,Energy,Homogeneity];
        mydataGLCM = [mydataGLCM, feat(:)];
        if k == 1
            save('newMethod.mat','mydataGLCM');
        else
            save('newMethod.mat','mydataGLCM','-append');
        end
    end

But this is storing features in a column-wise. 1 images feature are stored in 1 column. It should be row wise.

I think the question is clear now @ dpb

Thanks

[dpb]

"images feature are stored in 1 column."

Then transpose the column vector to a row vector with .'

[Sarajuddeen Nasrat]

hi dear when i want run your code i get error from matlab that is as follow: Cell contents reference from a non-cell array object. how can i correct it.

[dpb]

Don't use {} ("the curlies") to subscript what isn't a cell array.

[fadi ssuhimat]

don't use it, where?

[munipraveena rela]

mydataGLCM = [mydataGLCM; feat(:)];

stores data in row wise

[munipraveena rela]

mydataGLCM = [mydataGLCM; feat]

Answer 2

[fadi ssuhimat]

did you get the right answer... I have same problem now?