Build a temporal (unshared) data structure inside a compute shader

I'm working on a compute shader that has to internally create a binairy tree in order to calculate the value of several of the pixels values. Is there an easy way to create such a data structure with minimal overhead, I want this binairy tree to be easy to balance while reading in the data so that I can read about 1024 pixels and put them in without having to do overly large amounts of overhead.