An improved approach of lung image segmentation based on watershed algorithm

As a preprocessing step of chest Computed Tomography (CT) images, lung segmentation is significant for the diagnosis of lung disease. The traditional watershed algorithm is sensitive to the noise and has the drawback of over-segmentation problem. This paper presents a novel image segmentation method to improve Watershed segmentation algorithm with the maximum between-class variance algorithm (OTSU). We adopt the OTSU method and mathematical morphology method in the period of the initial image segmentation and then compute a segmentation function. Finally, we compute the watershed transform of the segmentation function. The experimental results point out that this method is an effective segmentation method of lung parenchyma, which lessens the problem of the over-segmentation in the lung image effectively and runs faster.