Extracrting high intensity objects from grayscale image

Question

My data is grayscale images taken from an IR camera. The objects I'm interested in have a higher intensity than the background. One problem is that the contrast between the object to its background changes from picture to picture. Another problem is that my object's value range can change. for example, the object can be in the range [170,255], and at other times in the range [70,255], or any other range. The only certainty is that the object will be with the highest intensity values in the image.

My goal is to extract the object and match it to familiar objects from my data set.

The first step I did is thresholding the images. The threshold is determined by the cumulative sum of the histogram of each image by creating a "cumulative histogram" and normalizing it. Next, I take the value of the x-axis that corresponds to the value 0.93 in the y-axis and that x value is my threshold.

Testing it on my data yields great results in filtering the objects, but still, there is some noise that doesn't get filtered out.

the images and the thresholded images

My question is what do you suggest to do in order to achive my goal from the point I'm at. I have thresholded images but from this point I can't extract exactly the object, it's always with noise, or if I try to filter the noise, some parts of the objects also get filtered.