基于高斯拟合的自适应多阈值分割算法  

Adaptive Multi-method Segmentation Algorithm Based on Gaussian Fitting

作　　者：李磊[1] 平西建[1] 童莉[1] 

机构地区：[1]信息工程大学信息工程学院,河南郑州450002

出　　处：《计算机应用研究》2006年第8期162-163,170,共3页Application Research of Computers

基　　金：国家自然科学基金资助项目(60272004)

摘　　要：提出一种新的基于邻域平均直方图的快速高斯拟合函数参数估计的算法,对形态滤波后的邻域平均直方图进行峰谷检出,对各峰区数据单独进行截断数据统计分析确定拟合参数,较好地避免了多峰交叠所带来的干扰,从而实现了复杂图像的自适应多阈值分割。实验表明,该算法能有效提高拟合精度和分割的鲁棒性。To alleviate these undesirable disturbance between peaks, and fit multi-peak feature of histogram more accurately, a new and fast estimation algorithm of Gaussian fitting functions＇ parameters based on local average histogram is introduced in this paper. After morphology filtering, peak and valley are detected in the histogram. Statistical theory for truncated data analysis are applied to estimate the parameters of fitting function of each peak area, and get optimum thresholds. Experimental results show that this algorithm can be automatically and effectively applied for the multi-threshold segmentation.

关 键 词：图像分割 邻域平均 高斯拟合 参数估计 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]