The accuracy and efficiency of the lung segmentation are significant to computer-aided detection/diagnosis (CAD/CADx) scheme for pulmonary nodules detection in chest computed tomography (CT) image. And morphology is widely utilized to characterize the shape of the object in lung segmentation. In this investigation, a multi-stages based approach which combines thresholding, connected component analysis and morphology is proposed to achieve a fast and precise lung segmentation. The presented framework consists of three stages: thorax extraction, lung segmentation and boundary refinement. A dataset of CT scans from different equipments and modalities is utilized to evaluate the proposed method. The average dice similarity coefficient (DSC) of the experiments is 0.97 and average time-consuming of each slice is 0.64s. The results demonstrate that the proposed method with multi-stages is an efficient and accurate method for lung segmentation.