Abstract: It is the purpose of the development of ferrography to analysis of ferrography images using the computer image dispose technology.The segmentation method of ferrographic debris image is the important tache，and its segmentation effect affects recognizing precision directly.The original region growing algorithm needs the supply of the initial point selection and the growing threshold.Different supply of these data will influence the segmentation effect seriously.A new region growing algorithm is presented.It can get the initial point selection and growing threshold automatically with the help of the FCM algorithm and the fuzzy mutual information.It implements the automatic segmentation of the wear deris image.Experimental results show that the algorithm is more feasible.

Key words: ferrography, wear debris, image segmentation, region growing, fuzzy mutual information

摘要： 利用计算机图像处理技术实现铁谱图像诊断自动化是铁谱技术发展的目标，铁谱磨粒图像分割是磨粒自动识别的重要环节，其分割效果直接影响磨粒识别的精度。原始区域生长算法需要提供种子点以及生长阈值才能进行图像分割。不同的种子点和不同的阈值会对分割效果产生很大影响。提出一种结合模糊C均值的区域生长算法，可根据磨粒图像自动获取种子点，并利用模糊互信息自动确定生长阈值，实现磨粒图像的自动分割。实验结果验证了该方法的有效性。

关键词: 铁谱, 磨粒, 图像分割, 区域生长, 模糊互信息