Abstract: The automatic segmentation of lung parenchyma is one of the key techniques to the computer-aided diagnosis (CAD) system for lung cancer, emphysema and other lung diseases. A new automatic lung segmentation based on the 3-D CT image series is proposed integrating the adaptive gray-level threshold, region growing and math morphological algorithms together. And the graph search algorithm is used at specified layers to position accurately the narrow zone formed by the anterior and posterior lines connecting the left and right lung, thus overcoming efficiently the difficulties that the nodules at lung edge are easy to be missed and how to separate the left lung from right one. The experimental results of many sets of CT images verified that the technique proposed is highly exact and efficient for lung parenchyma segmentation.