Abstract: To segment accurately the pathologic lung images, the conventional segmentation algorithm for lung was analyzed and improved by introducing the wavelet transform and mathematical morphology into the algorithm, i.e. an image was decomposed via wavelet transform, then each and all of its components after decomposition were mended in different ways through mathematical morphology. Morphologically the closed and open operation should be done for LF and HF components, respectively, thus modifying the basic traits of an image on a proper scale without detailed traits obscured and, after image restructuring, obtaining the ideal lung region. The improved algorithm was applied to the segmentation of 36 sets of clinical HRCT data provided by hospitals, and its results were compared with those from manual segmentation and conventional algorithm. It was found that the improved algorithm can provide accurate segmentation so as to make the average sensitivity of lung segmentation higher.