standard deviation of image

I have a 3-dimensional image (image1) and a 2-dimensional image (image2) with values from 1 to 7 and want to find the standard deviation for specific values. I'm using the following code to calculate the mean value but dont know how to calculate the standard deviation. Any suggestion please?

for i=1:3

x=image1(:,:,i)

for j=1:7

L = bwlabel(image2==j);

STATS = regionprops(L,x,'MeanIntensity','Area','PixelValues');

n_stats = size(STATS,1);

   area_=zeros(n_stats,1);

   mean_=zeros(n_stats,1);

[m1,n1] = find(cat(1,STATS.Area) >=5);

for ii=1:n_stats

      mean_(ii)=STATS(ii,:).MeanIntensity;

 end

mean1=mean_(m1);

mean2(j,i)=mean(mean1);

end

Answer 1

[Sean de Wolski]

You can use regionprops to calculate the various fields requested of each label in a label image, e.g.:

A = uint8(rand(10)*10); %random label image
STATS = regionprops(A,'area');
areas = [STATS(:).Area]

And I don't understand what you want the standard deviation of. If youu want it of all objects with the same value, won't it be zero? Clarify your goal, I guess. (and look at stdfilt)

[Walter Roberson]

Ah, that confused me as well, but I now see that the "2-dimensional image with values from 1 to 7" does not necessarily imply that the original image is solid color in those areas. Each of the areas in the 2D image acts as a mask to be applied to the 3D image.

[Image Analyst]

No it won't be zero because he's passing in the (badly named) original image x, so the pixel values of the blob will be those of the original image, not the labeled binary image.

[Hassan]

Thanks Sean for the comment. As I said I have 2 images, image 1 is a 3-dimensional image and image 2 is a categorised 2-dimensional image with values 1 to 7. I want to find the standard deviation of the pixels values from image 1 that has a value of 1 (and then 2, 3,4 ,..,7) in image 2.

[Hassan]

Sorry guys for not desribing the problem clearly. You are right Walter 'Each of the areas in the 2D image acts as a mask to be applied to the 3D image.'

thats right Image Analyst.

Answer 2

[Image Analyst]

I don't have MATLAB on this computer so I can't test but I believe it would be something like

pixelValues = [STATS(k).PixelValues];
sd = std(pixelValues(:)); % StDev of kth blob

or something like that.

[Hassan]

What is 'k' in STATS(k)?

[Image Analyst]

That's the kth region. You'd have it in a loop over all k, all possible blob numbers, like

[labeledImage, numberOfBlobs] = bwlabel(binaryImage, grayImage, 'PixelValues');

sd = zeros(1, numberOfBlobs);

for k = 1 : numberOfBlobs

pixelValues = [STATS(k).PixelValues];

sd(k) = std(pixelValues(:)); % StDev of kth blob

end

[Hassan]

I couldnt run the code. I got the following error messageç

??? Error using ==> bwlabel

Too many input arguments.

I think binary image should be the 2'dimensional categorized image and gray image should be the 3'dimensional image.

[labeledImage, numberOfBlobs] = bwlabel(image 2, image 1, 'PixelValues');

sd = zeros(1, numberOfBlobs);

for k = 1 : numberOfBlobs

pixelValues = [STATS(k).PixelValues];

sd(k) = std(pixelValues(:)); % StDev of kth blob

end