A method was devised to computationally segment and measure three-dimensional pulmonary trees in situ. Bronchi and pulmonary vessels were computationally extracted from volumetric computed tomography data based on radiopacity differences between airway wall and airway lumen and between blood and parenchyma, respectively. The tree was reduced to a central axis to facilitate measurement of branch segment length and angle. Cross-sectional area was measured on a reconstructed computed tomography slice perpendicular to this central axis. The method was validated by scanning two Plexiglas phantoms and an intact lung. Reconstructed diameters in the phantoms were accurate for branches > 2 mm. In the lung airway branches between 1 and 2 mm in diameter were often unresolved when their angle of orientation with respect to the axis of the scanner was > 45 degrees. However, if a branch was resolved, its reconstructed diameter was little affected by orientation. This method represents a significant improvement in the analysis of complex pulmonary structures in three dimensions.