— The current access point (AP) association schemes in wireless LANs, such as IEEE 802.11, cause an unbalanced load which reduces the performance of both the entire network and individual users. Intensive studies were motivated to determine efficient methods of balancing loads among APs. Previous works provided either the heuristics without theoretical analysis on the performance, or algorithms that require a centralized node or the propagation of global information. In this paper, we model the AP association problem as the many-to-one matching problem in the bipartite graph. Our objective is to maximize the total load among all APs. We propose a localized algorithm that provides a bounded approximation ratio in terms of total load and does not require the propagation of local information. We also extend the localized algorithm through iterative executions, and allow for adaptation to various environments. Extensive simulations are conducted to verify our results.