【摘要】 传统的模糊C均值(FCM)聚类算法广泛用于图像的自动分割,但它有两个缺陷:一是收敛速度过慢;二是当图像的目标和背景像素拥有相近的灰度值,具有相似的隶属度,导致了图像边界区域的不连续和模糊.针对该问题,提出一种改进的算法,在快速FCM聚类的基础上,利用粗糙集理论中的上近似和下近似的概念来描述图像的目标和背景,引入粗糙熵的概念,选择合适的阈值,对图像进行精确分割.实验结果表明,这种算法可以达到满意的分割效果.更多还原

【Abstract】 The conventional Fuzzy C-Means(FCM) clustering algorithm was widely used in automatic segmentation of image.However,there were two defects: one being slow converging speed;the other being that the boundary regions of image would become discontinuous and vague when the image’s object pixels and background pixels had similar grey scale and membership grade.Therefore,a improved algorithm was proposed to eliminate them.On the basis of fast FCM clustering,the object and back ground of the image were described with the concepts of upper and lower approximation in the rough sets theory and a concept of rough entropy was introduced.Therefore,the image would be segmented accurately after a threshold value being chosen.更多还原