FCM聚类算法和粗糙集在医疗图像分割中的应用

Medical image segmentation based on FCM clustering and rough set

本文通过结合FCM聚类算法和粗糙集,提出了一种新的图像分割方法。首先,以不同聚类数情况下FCM的分割结果为依据构建属性值表,基于属性构成的不可分辨关系将图像分成多个小区域;然后,通过值约简获得各属性权值并以此为依据,计算各区域之间的差异度,进而通过差异度定义的等价关系,实现各区域相似度评价;最后,通过相似度定义的最终等价关系实现区域合并,完成图像分割。该方法在人工生成图像和大脑CT图像及MRI图像的分割中得到验证,实验结果表明,本文方法和FCM方法相比,可以降低错分率,且对模糊边界区域的分割效果较好。

A new image segmentation method is introduced,which integrates FCM clustering and rough set theory. First,using FCM clustering results with different cluster numbers, an attribute value table is constructed; and the image is divided into many small regions based on the indiscernibility relations of the attributes structure. Then,using ‘value reduction＇, the weights of the attributes are obtained,and the dissimilarities of the regions are calculated. The equivalent relations defined by the dissimilarities are used to evaluate the similarities of the regions. Lastly, the final equivalent relations based on the similarities are used to implement the combination of the similar regions and the image segmentation is accomplished. The method was applied to an artificial generated image, a human brain CT image and a MRI image. The experimental results indicate that, comparing with FCM method, the proposed method decreases the error rate of classification and has better segmentation performance for vague boundary regions.