We present a new method for computerized tomography (CT)-based investigation of pulmonary airways. The developed approach is based on a fully automated 3-D reconstruction of the bronchial tree from volumetric spiral CT data. The 3-D reconstruction procedure combines mathematical morphology-based 2-D segmentation, performed on axial images, together with 3-D topology-based restoration and filtering. The 3-D bronchial tree is reconstructed up to the sixth subdivision order and is investigated by using a volume-rendering technique, where bronchi voxels are assigned the gray-level value of their projection in the original CT images. Such a visualization modality, referred to as CT bronchography, provides radiologists with a perception of the sole 3-D bronchial tree structure and preserves the original data contrast. Tested on a data set corresponding to 20 clinical examinations, the proposed method shows robustness with respect to various pathologies such as stenosis, bronchiectasis, mucoid impactions, and infiltrative diseases. It emphasizes the interest in considering CT bronchography as a complementary tool for clinical diagnosis and follow-up of airway diseases.