资源描述：
《指纹论文：指纹图像分割方法研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、指纹论文：指纹图像分割方法研究【中文摘要】指纹分割作为指纹识别的预处理环节,不仅能提高指纹特征提取精度,而且能减少指纹预处理时间,对提高整个识别系统性能有着重大意义。本文在对常用指纹分割方法进行分类分析和探讨的基础上,对指纹图像分割理论和技术进行了深入研究,并对现有分割算法进行了改进。本文主要工作包括:(1)对指纹图像分割方法做了比较全面的分析综述。从指纹特征提取角度出发,将指纹图像分割方法分为基于特征融合判决分割方法与基于多级分割思想分割方法两类,并对这两类算法进行了详细分析。(2)针对特征融合判决分割方法,把最小平方误差准则用于基于线性分类的指纹图像分割算法,该算法对低质量
2、指纹图像分割效果较好;将指纹图像纹理特征引入灰度方差求解过程,提出一种基于纹理特征的指纹图像自适应分割算法。通过实验对新算法分割效果及噪声抑制能力进行了验证。(3)针对多级指纹分割方法,研究了结合多种方法的高效指纹图像逐级分割算法,采用一种鲁棒性更好的求点方向图方法,计算前景块中各像素方向;采用自动确定部分阈值的分级分割方法,改善算法中仅凭经验设定多个阈值的缺点。实验结果表明,新算法分割精确率较高。【英文摘要】Asanimportantstepoffingerprintimagepreprocessing,fingerprintsegmentationwhichhasagrea
3、tsignificanceinthesystemperformancecannotonlyimprovetheaccuracyofthefeatureextraction,butalsoreducethetimeoffingerprintpreprocessing.Thispapermainlyinvestigatesthetheoryandmethodsoffingerprintimagesegmentation,andsomeofthesegmentationmethodsareclassifiedanddiscussed.Themainworksinclude:(1)Ac
4、omprehensivereviewofthefingerprintimagesegmentationmethodsisgiven.Basingonextractingfingerprintfeatures,wefirstclassifythemethodsoffingerprintimagesegmentationintotwoclasses,whicharethefeaturefusionsegmentationmethodandmulti-levelsegmentationmethod.Andthen,weintroducethetwomethodsindetail.(2
5、)Intheaspectofthefeaturefusionbasedfingerprintsegmentationmethod,weinvestigatetheMinimumSquareErrorrulebasedfingerprintsegmentationalgorithmandaadaptivetexturefeaturebasedfingerprintsegmentationalgorithm.First,weinvestigatethelinearclassifierbasedfingerprintsegmentationalgorithm,whichusesaco
6、mbinationofthreevariance,meanandridgeorientationfeaturesoffingerprintimage,andpresentstheimprovedalgorithmofusingtheMinimumSquareErrorrule.Theexperimentalresultsdemonstratetheeffectivenessoftheimprovedalgorithm,especiallyinlowqualityimages.Second,weinvestigatethetraditionalvariancebasedfinge
7、rprintsegmentationalgorithm,andpresentaadaptivetexturefeaturebasedfingerprintsegmentationalgorithmwhichcombinetexturefeatureandvariance.Theexperimentalresultsdemonstratetheeffectoftheproposedsegmentationalgorithmisbetterthanthevariancebasedfingerpr
