摘要
根据石油地质部门的实际需要,提出了一种改进的ISODATA聚类算法,并以此对彩色荧光图像进行分割.该算法首先从统计模式识别的角度出发,将彩色图像各个像素点的色彩分量转换到更符合人视觉的HLS空间;并在进行自组织迭代前,对样本空间进行初分,为取得更好的聚类效果创造了条件;最后,在进行相似度运算时,根据不同的分割需要分别对色度、亮度、饱和度三个分量进行了加权处理,获得了较好的分割效果.
The article puts forward an improved ISODATA Algorithm which is used to segment the color fluorescence image for the oil geological department. The algorithm converts the color from the RGB space to the HLS space which is fit for human's vision, then does a pre-segmentation before ISODATA process to make a better clustering result. During the similitude calculation, the algorithm gives different weights to each of the three scalars which are Hue,Lightness and Saturation,to meet different segmentation needs, and improve the clustering result too.
出处
《四川大学学报(自然科学版)》
CAS
CSCD
北大核心
2007年第3期563-568,共6页
Journal of Sichuan University(Natural Science Edition)
基金
四川省攻关项目(05GG021-026-03)
作者简介
何霁(1981-),男,2004级硕士研究生.
通讯作者:罗代升.E-mail:cias@263.net.