基于二维灰度直方图Fisher线性分割的图像分割方法  

THE GRAY-LEVEL THRESHOLDING METHOD BASED ON FISHER LINEAR DISCRIMINANT OF TWO-DIMENSIONAL HISTOGRAM

作　　者：龚坚[1,2] 李立源[1,2] 陈维南[1,2] 

机构地区：[1]东南大学自动化研究所,南京210018 [2]中国科学院自动化研究所国家模式识别实验室,北京100080

出　　处：《模式识别与人工智能》1997年第1期1-8,共8页Pattern Recognition and Artificial Intelligence

摘　　要：本文分析了图像二维灰度直方图上聚类分布的特征,推导出了二维灰度直方图的Fisher线性判别函数,分析表明,相对于现有直接在二维灰度直方图上进行向量分割的方法,新方法不仅能降低误判的概率,同时,将二维问题转化为一维问题,大大减少了计算量.实验表明,新方法比直接利用一维和二维灰度直方图进行阈值分割的方法均优越.In this paper, the features of the clusters on the 2-dimensional (2D) histogram of the image are analyzed, and then the Fisher linear discriminant of the 2D histogram is derived. Analysis shows that as to the current methods of separating 2D histogram with a threshold, the new method which uses the Fisher linear discriminant to segment image can not only reduce the error rate caused by noise, but also decrease the computational time greatly by converting the 2D problem to the 1-dimensional (1D) problem. Experimental results show that the performance of the new method is better than that of those simply using criteria based on 2D histogram.

关 键 词：图像分割 阈值化 FISHER线性判别 图象识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]