how to find the number of atoms at different peak positions automatically?

I have a set of maximum positions and there are some number of atoms at each position. I get the number of atoms by manually pointing at each peak position. But how can I find a way that automatically counts the number of atoms at each max position?

[Image Analyst]

[sowmya]

I have another question, but not related to it I guess. I have a microsocopy image which i am attaching. I have both .tif and .mat file for the image. I want the positions of all the bright spots there which correspond tot he atoms.

I think i should fit gaussian to find the peaks. Please comment and help.

Thank you!

[Image Analyst]

Find each individual spot, find it's centroid and second central spatial moment (demo attached), then use fspecial().

Answer 1

[Image Analyst]

You could try findpeaks() if you have the Signal Processing Toolbox.

Attach 'cube.mat' and a screenshot of the signal if you need more help.

[sowmya]

The Cube file exceeds the size limit. I do not kow how to further compress it externally. I am attaching the screenshot of the end result i.e., the projection_img.

I already have the information about peaks and its positions, so I dont have to use peakfinder I think. am I right?

now I want to count atoms at each peak position without i having to place the cursor at each point.

[Image Analyst]

Finding the spots is easy enough - you can use thresholding and regionprops(), or imregionalmax() - but once you've found a peak location, how do you turn that (it's size, peak value, or integrated gray value) into a number like 18?

[sowmya]

As you can see in the code attached, after finding the spots, value 1 is assigned to each of the positions and summed along the direction to get the number of atoms. let me know if it is right and makes sense?

[Image Analyst]

What I would do is to threshold, then call

props = regionprops(mask, 'WeightedCentroid');

Then get the y locations

xy = vertcat(props.WeightedCentroid);
y = xy(:, 2);

Then you need to count the number in each "Band" but the spots may not be perfectly aligned. So you can take a mean across the image and threshold then use regionprops() again this time to find the centroid of the black bands in between the spots bands.

meanVerticalProfile = mean(mask);
binaryImage = meanVerticalProfile < someDarkThreshold;
props2 = regionprops(binaryImage, 'WeightedCentroid');
xy = vertcat(props.WeightedCentroid);
edges = xy(:, 2);

etc.

Now you can use those black band centers as "edges" in histogram() or histcounts() to get the count in each spots band

counts = histcounts(y, edges);

It's all untested since I don't have your data, so adapt it if it doesn't work.

[sowmya]

Okay. I will try it out. Thank you!