Lung disease has risen to the third leading cause of chronic morbidity and mortality in the United States. The diagnosis, differentiation, and classification of the severity of various lung diseases rely on clinical assessment, thoracic imaging using computed tomography (CT), and pulmonary function testing (PFT). While being the reference standard for assessment of the lung's mechanical function, PFT strictly permits a global measurement of lung physiology. In contrast, high-resolution image analysis is a powerful tool with the potential for regional as well as global quantification of diseases. Imaging plays an increasingly important role in lung disease diagnosis. Most current pulmonary imaging techniques are used clinically to assess anatomic changes and to provide qualitative or semi-quantitative estimates of disease severity. Although generally effective, radiologic interpretation of CT images is time-consuming, requiring considerable expertise. It is also largely qualitative and prone to inner-observer diagnostic variability. It is therefore desirable to have automated quantitative analysis from imaging modalities. This dissertation investigates the problem of using thoracic computed tomography images for automatically and quantitatively analyzing and diagnosing lung diseases. We discuss how to complete a holistic automatic pipeline for clinical studies and study its various components. The final goal is to illustrate how to build a complete pipeline using CT images as input and yielding the desired clinical results as output. In the case when a single lung image is provided, we propose an algorithm for segmenting small airways from CT images. When the images at both inspiration and expiration phases are available, we discuss the use of image registration algorithms to compute lung kinematics, especially focusing on various diffeomorphic transform models. Finally, these image-derived quantitative metrics are analyzed for clinical studies of specific diseases, including differentiation of interstitial lung diseases and chronic obstructive pulmonary diseases, and quantification of small airway air trapping and emphysema. We show that quantitative CT imaging, integrating segmentation, registration, feature computation, feature selection and pattern recognition, can provide better biomarkers for diagnosis and prognosis.