Segmenting objects from a threshold image

I have thresholded an image using thresholder app in Matlab and converted the output for an image to black and white. AS a result I got the following output,

Now what I need to do is to isolate the specific type of square in the thresholded image from above image: I want my output to be as follows

I have tried bwareafilt(mask, [minArea maxArea]); command but that dosen't give optimal solution, more on that is here https://www.mathworks.com/matlabcentral/answers/676728-how-to-implement-helping-functions-for-counting-lego-blocks-in-image

Please follow comment section in above link to understand the problem of using bwareafilt command.

Is there anyone who knows solution for how to tackle this problem.

Thank You in anticipation.

[Image Analyst]

What is it about that particular blob that distinguishes it from any other blob? Like do you always want the smallest blob? Or a blob in a particular (row, column) location? Or a blob with a particular mean intensity of the corresponding gray level image? How did you know that you want that particular blob instead of any of the others???

If you need more help, can you post the gray level image and the segmentation code?

[Adnan Khan]

It is the size that distinguishes it from others. It can be placed anywhere in other images. Below is the actual image, which was first thresholded using thresholder app, and the result generated from it is changed into black and white.

here I am not posting the code which was auto generated by matlab thresholder app; the function's name used for thresholding blue blobs is createMaskBlue() which efficiently threshold's the blue blobs.

code is:

[maskBlue, maskedRGBIm] = createMaskBlue(I);

maskBlue = bwareafilt(maskBlue, [26000 75000]);

% Count number of Red blobs

[labeledImage, numB] = bwlabel(maskBlue);

imshow(labeledImage)

The labeled image is the first one posted in question

and rgb image is the following

[Adnan Khan]

The blue blobs here are of different sizes as you may see in rgb image. Based on the size, I want the smallest blob's count, not only for this image but for more such images also, where the smallest blob can be placed in different positions. I hope my question is clear now.

Answer 1

[Image Analyst]

"Based on the size, I want the smallest blob's count" <== If by count you mean area in pixels of the smallest blob in the segmented image, you can do this:

smallestBlob = bwareafilt(maskBlue, 1, 'smallest');
areaInPixels = nnz(smallestBlob)

[Adnan Khan]

if there are two same sizes blobs present in an image but with different orientations, in that case it will output only one of them and not both. I want something concrete that extracts the dimensions of the blob with certain flexiblity and ensures to catch more than one blobs even if they are in different orientations. Shall I tune this '1' value for that? is it for outputting just one blob?

[Adnan Khan]

and by count I mean, to count number of blobs present in the image of the smallest size.

[Image Analyst]

With an image that is millions of pixels, it's highly unlikely that you will have two blobs with the same area. If you see that happening, I'd like to see the original image and mask image. However there is a way to do it.

% Measure areas of all blobs:
props = regionprops(mask, 'Area')
% Get all the areas into a double vector
allAreas = [props.Area];
% Find the smallest area, in pixels.
minArea = min(allAreas)
% Count the number of blobs that have the minArea exactly.
blobCount = sum(allAreas == minArea)  % Will be 1 in almost every case unless you've synthetically created an image with known areas.