The image I used for the ImageJ analysis was of a group of flamingos.

I converted the image to 8 bit and set the threshold at 2 different levels to attempt to get the best count.  An issue was that some of birds were very close together and were counted as one.  I did not exclude edges since there are birds on the edges, but ones that are mostly off the photo were not counted.  The count is going to be underestimating the actual number due to these factors.  I attempted to set the threshold to separate them as much as possible without losing any of the birds, used a pixel range of 5-infinity with circularity set to a minimum of 0 (since they are not round at all), and the best count was 304.

Data:

Slice Count Total Area Average Size Area Fraction Mean Major Minor Angle
flamingo.jpg 304 3535 11.628 4.3 254.958-
flamingo.jpg 297 3368 11.34 4.1 254.957-