What you need is an ‘order-statistics tree’. It is essentially an augmented (binary search) tree that supports the additional operation rank(x) which gives you the number of elements with less or equal key as element x. Chapter 14 in Cormen, Leiserson, Rivest, Stein; “Introduction to Algorithms” should give you the algorithmic background.
There is also some implementation on the web.