融合颜色和区域信息的彩色图像分割新方法  

Color image segmentation method using color,spacial and texture inforation

作　　者：潘红艳[1] 敬忠良[2] 肖刚[2] 郭超[1] 

机构地区：[1]上海交通大学电子信息与电气工程学院自动化系,上海200240 [2]上海交通大学空天科学技术研究院,上海200240

出　　处：《计算机工程与应用》2007年第25期72-74,共3页Computer Engineering and Applications

摘　　要：结合了均值漂移算法和区域合并算法,取长补短,提出了一种融合颜色和区域信息的彩色图像分割方法。该算法首先利用均值漂移求取各个局部极值(聚类中心),在带宽求取和权重设置上使用了自适应法则,使算法更具有适用性;然后使用一个基于阀值的区域合并算法,解决了均值漂移对纹理和关照变化的过分割。实验证明,该算法是有效的。Color image segmentation is a critical step from color image processing to image analysis.A new color image segmentation method using fused information of color and space features have been proposed by combining mean-shift and region merging.It can complement their strength and weakness.An ameliorated mean shift method is used to gain local minimums,namely the cluster center,with the bandwidths and weights acquired according to self-adaptive principle,which makes the algorithm more applicable;then the over-segmentation of texture and illumination caused is settled by valve based region merging.The results demonstrate that the method is effective and optimistic.

关 键 词：均值漂移 彩色图像分割 区域合并 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]