GLCM stats with a mask

Edit question

I would like to evaluate the glcm features of the region of interest that in inside the mask. To only consider what is inside the mask I am attempting to multiply the original image by the mask and then take the glcm. I keep getting the same error.

I = rgb2gray(imread('5.jpg'))
imshow(I)
hFH = imfreehand()
xy = hFH.getPosition;
binaryImage = hFH.createMask();
blackMaskedImage = I;
blackMaskedImage(~binaryImage) = 0;
binaryImage=blackMaskedImage*blackMaskedImage(~binaryImage)
glcm_binaryImage=graycomatrix(binaryImage)
imshow(blackMaskedImage);
stats_binaryImage = graycoprops(glcm_binaryImage)

[Walter Roberson]

What is the error that you get?

[Donnell Perkins]

Answer 1

[Donnell Perkins]

%%Working Code

fontSize = 20;I= rgb2gray(imread('C:\Users\x\Desktop\5.jpg'))

subplot(1, 2, 1);

imshow(I);

title('Original Image', 'FontSize', fontSize);

subplot(1, 2, 2);

imshow(I);

title('Draw Region Of Interest', 'FontSize', fontSize);

set(gcf, 'units','normalized','outerposition',[0 0 1 1]); % Maximize figure.

set(gcf,'name','Demo by ImageAnalyst','numbertitle','off')subplot(1, 2, 2);

hFH = imfreehand()

maskImage = hFH.createMask();subplot(1, 2, 2);blackMaskedImage = I;

blackMaskedImage(~maskImage) = NaN;imshow(blackMaskedImage);title('Masked Image', 'FontSize', fontSize);maskedImage = I.* cast(maskImage, class(I));glcm_maskedImage=graycomatrix(maskedImage)

stats_maskImage = graycoprops(glcm_maskedImage)

Answer 2

[Walter Roberson]

I = rgb2gray(imread('5.jpg'))
subplot(1,2,1)
imshow(I)
hFH = imfreehand()
xy = hFH.getPosition;
binaryImage = hFH.createMask();
blackMaskedImage = double(I);
first_unused = max(blackMaskedImage) + 1;
blackMaskedImage(~binaryImage) = first_unused;
subplot(1,2,2);
h = imshow(blackMaskedImage);
h.AlphaData = blackMaskedImage ~= first_unused;
drawnow();

glcm_binaryImage = graycomatrix(blackMaskedImage);
glcm_binaryImage = glcm_binaryImage(1:first_unused-1, 1:first_unused-1);

stats_binaryImage = graycoprops(glcm_binaryImage)

When you use the masks the way you were using them, you were introducing pixels that were 0, which resulted an more 0's being present in the image then there should be. graycomatrix was not going to ignore the 0s because the 0's are valid pixel values.

This code that I present changes the masked entries to a pixel value that would otherwise not exist. With that in hand you can calculate the co-occurance matrix, and then throw away the parts of the co-occurance matrix that relate to the mask pixel value.

I carried over the variable names you were using, but they are inaccurate in several cases. Your line

blackMaskedImage = I;

does not create a binary image: it copies the values from I, whatever data type they were, very likely uint8 for .jpg . If you think that your image is a binary image, then chances are that really it contains values 0 and 255, not 0 and 1, and your co-occurance matrix would include that entire 0 to 255 range. But as an outsider, I must assume that you are starting with an image with multiple shades that should not be binarized, becuase you did not binarize it.

[Donnell Perkins]

Thank you so much for explaining my error. I understand what I did wrong now. But I am still continuing to get an error when I try to define anything with (~binaryImage) in it. It has happened to me several times while troubleshooting as well

[Walter Roberson]

I should have used

first_unused = max(blackMaskedImage(:)) + 1;