regionprops and standard deviation (binary image)

Hi everyone, i am using the function" regionprops" to get the largest perimeter in my binary image. I have got the value of the largest perimeter but still need to have its standard deviation. Does anyone know how to get this std? Thank you very much

Answer 1

[Walter Roberson]

When you do the regionprops() also ask for 'PixelIdxList' . When you find the region with the largest perimeter, extract its PixelIdxList. Use that PixelIdxList to index the 2D image, and take std() of the pixels that are returned.

If your original image was color, then sometimes the easiest is to break it into 3 variables and use PixelIdxList on each. That is not the only way, but it is easy to understand whereas the conversion of 2D linear indices to appropriate 3D indices can be a bit tricky to understand.

[marie semaan]

I appreciate your help in this matter.

I have written the following code:

s = regionprops(binaryimage,'PixelIdxList','Perimeter')% original image is not a color image

allPerimeters =[s.Perimeter]% all perimeters in this binary image

largestPerimeterIs=allPerimeters(1); % the largest perimeter value is the first value in allPerimeters 

for k = 1:NumObjects
    s(k).RegionNumber = mode(single(originalImage(s(k).PixelIdxList(1))));
 end
region_vector = [s.RegionNumber]

How can i extract the PixelIdxList of the region with the largest perimeter in order to apply the std of all those returned pixels?

Thank you again

[Walter Roberson]

originalImage(s(k).PixelIdxList(1)) is going to be a scalar value, so I do not understand why you would mode() it?

"How can i extract the PixelIdxList of the region with the largest perimeter"

[~, objidx] = max(allPerimeters);
std( double(OriginalImage(s(objidx).PixelIdxList)) )

[marie semaan]

by trying the following code, std is equal to 0. What shall i change in the "std(double(OriginalImage(s(objidx).PixelIdxList)))"?

s = regionprops(binaryimage,'all')
[~, objidx] = max(allPerimeters);
std(double(OriginalImage(s(objidx).PixelIdxList)))

thank you

[Walter Roberson]

s = regionprops(binaryimage,'all')
[~, objidx] = max(allPerimeters);
obj_pixels = OriginalImage(s(objidx).PixelIdxList);
if length(unique(obj_pixels)) == 1
  sprintf('all pixels in the object are exactly the same! All of them have value %d\n', obj_pixels(1));
else
  std(double(obj_pixels))
end

Depending how your image was constructed and how you are constructing your binary image, the situation could either be an improbable coincidence or something to be expected. You could post your image and your code that creates the binary image for us to examine.

[marie semaan]

Thank you

Answer 2

[Image Analyst]

You can get the standard deviation of the gray levels in the blob with the largest perimeter more directly (than by using PixelIdxList) like this:

% Get blob with largest perimeter.
binaryImage = bwpropfilt(binaryImage, 'Perimeter', 1);
% Get standard deviation of pixels in that blob
stdGrayLevels = std(double(grayImage(binaryImage)))

[marie semaan]

Thank you for your help. I prefer to use the above code in my case. i have tried your code but it did not work.