摘要
改进了基于地形学距离的分水岭算法,提出了一种结合了图像灰度、边缘信息与信息熵的图像分割方法。首先利用改进的分水岭算法将图像分成多个小区域,根据各个区域之间的临接关系,建立RAG;其次,利用提出的区域相似度合并区域;最后根据最大熵准则停止合并过程,获取最终的分割结果。实验结果表明,与改进前的分水岭算法相比,该方法边缘定位更加准确。与k-mean和基于边缘的分割方法相比,能够较好地分割出图像的细节,同时分割结果也更加符合人的视觉特性。
The paper proposed a improved algorithm of watershed by topographical distance and a scheme for image segmentation based on grayscale ,edge information and information entropy, First ,an image is separated into a large number of small partitions by a improved watershed algorithm and a RAG is built according to the adjacent relationship among partitions ; second , region merging is performed on the basis of a region similarity ; finally, the maximal total entropy criterion is used to stop region merging and gain the final segmentation result. The result show that comparing with a original watershed algorithm , the edge position of the new method is more accurate and the image segmentation result of the proposed approach is more consistent with human vision properties with more detail information of the image than that by the k - mean and a technique based on edges,
出处
《微计算机应用》
2007年第11期1132-1137,共6页
Microcomputer Applications
关键词
图像分割
分水岭
区域熵
区域合并
image segmentation, watershed,region entropy, region merging
