This paper addresses the relay assignment problem in a wireless cooperative network. We propose a distributed algorithm which does not require global information, hence is more practical in a real network environment. It is based on the distributed auction structure that can achieve the optimal network capacity as centralized ones but with much lower complexity and higher scalability. We give formal proofs for optimality and convergence of this algorithm, and then we analyze its time complexity and further improve it by reducing the iterations in cost of little capacity degradation. Extensive simulations also show huge computation reduction on proposed algorithms than centralized ones with comparable capacity performance.