There are various challenging options to accelerate parallel rendering through improved compositing such as: optimization of the compositing pipeline, on-the-fly compression or variable sub-resolution rendering. However, distributed parallel compositing is a complex construct which has to balance various aspects. Overall the major challenge is to transfer only the minimal amount of data as fast as possible and doing so with negligible overhead in computing cost. Also, the data reduction and transmission must be generic to support general purpose parallel rendering. Thus we develop flexible distributed image compositing that can adjust to various rendering scenarios.