How to select very small components in an image

Question

I have been trying to identify the centroids or positions of the small yellow dots (not the large regions) in the image below. I have adopted the following strategy (please see the code), however, i wonder if the results can be improved. In my attempt, I could not identify all spots and rather some extraneous spots are also selected. Thanks !

The original picture:

The picture with the centroids of the spots highlighted, is as follows:

The code can be found below:

segmented = ColorSeparate[img][[2]] // MorphologicalPerimeter[#, 0.2] & // MorphologicalComponents //
SelectComponents[#, {"Count", "AdjacentBorderCount","Holes"}, #1 < 50 && #2 == 0 && #3 >=  0 &] & 
// Colorize

centroids = ComponentMeasurements[segmented, "Centroid"];
HighlightImage[img, ListPlot[centroids // Values]]

Also do you recommend the use of Binarize[img] in the very beginning?

Answer 1

I will recommend a function of TopHatTransform for this request,you wil get it like this:

mask = MaxDetect[TopHatTransform[img, DiskMatrix[1]], .2]

Show it in your original picture:

HighlightImage[img, {Gray, mask}]