摘要
有效的指纹分割能够提高特征提取精度和减少后续处理时间,针对这种情况,提出一种基于D-S证据理论的指纹图像分割算法。该算法首先用改进的灰度方差和均值进行初级分割,然后进行平滑;其次使用方向性和对比度两个信息分别作为两个分类器的特征,并利用模糊规则对各分类器的基本概率分配判断,最后利用D-S证据理论的合成法则将两个分类器的结果进行融合判决,实验结果验证了算法的有效性。
Effective fingerprint segmentation can improve the accuracy of feature extraction and reduce the follow-up processing time.In view of this situation,a segmentation method is proposed for fingerprint image based on D-S evidence theory.This method firstly uses improved mean and variance to remove background,and then smooths it;Secondly,it uses contrast and orientation coherence as features for two separate classifiers based on the characteristic of the fingerprint images.The feature extracted from the image is usually uncertain.And using fuzzy rules makes a decision for classifiers base probability distribution.Lastly,the D-S evidence theory is used to determine the final result of two classifiers,experimental results verify the feasibility of this algorithm.
出处
《计算机工程与应用》
CSCD
北大核心
2010年第24期169-172,共4页
Computer Engineering and Applications
基金
安徽省教育厅自然科学研究项目(No.KJ2009B136)~~
关键词
图像分割
D-S证据理论
方向强度
对比度
image segmentation
D-S evidence theory
orientation intensity
contrast
