Lots of ways, like measuring all blobs and looking for which have a circularity near 1. Or extract the 2 biggest blobs, then take the smallest of those two (see attached demo). Plus other ways depending on what assumptions you make.
See my Image Segmentation Tutorial in my File Exchange to see how to do filtering based on size. You could extend it to filter based on size and circularity (= perimeter^2/(4*pi*Area)).
