Causal inference plays an important role in decision making in many fields, such as social marketing, healthcare, and public policy. The major challenge in estimating causal effects with observational data is to handle the confounding effects. Propensity-score based matching and reweighting methods are frequently adopted when multiple confounders need to be controlled. In practice, researchers are inclined to consider as many as potential confounders in order to fully adjust the selection bias. The performance of the propensity-score based approaches may be unsatisfactory in the high dimensional scenarios. Another stream of research for controlling the confounding effects is to directly balance the distributions of the confounding factors in the treated and control groups. In this work, we propose an algorithm to jointly select confounders and balance the selected confounder distributions. We conduct extensive experiments on both synthetic and real datasets. The experimental results demonstrate that our proposed method outperforms the state-of-the-art methods in many high dimensional settings.