QQ客服 官方微博 反馈留言

题名	生物信息处理:对自动指纹识别和医学图像分割的研究
作者	罗希平
学位类别	工学博士
答辩日期	2001-06-01
授予单位	中国科学院研究生院
授予地点	中国科学院自动化研究所
导师	田捷
关键词	自动指纹识别 医学图像分割
学位专业	模式识别与智能系统
中文摘要	自动指纹识别系统（Automated Fingerprint Identification Systern，简称AFIS） 有着广泛的应用背景。指纹图像增强和细节匹配是AFIS的两个重要问题。本文给 出了一种基于方向场的指纹图像增强算法，并对Anil Jain等人提出的细节匹配算 法进行了修正。我们采用了一种新的更简单的方法来进行指纹图像的校准，并以 一种简单而有效的方式将脊线信息引入匹配过程中，这样做的好处之一是以较低 的计算代价有效地解决了匹配中参照点对的选取问题。另外，我们采用了大小可 变的限界盒来适应指纹的非线性形变。在FVC2000公布的指纹图像数据库上按 FVC2000测试标准所做的实验显示我们的算法对原算法有较大的改进，且有较快 的速度和较高的准确率。 图像分割是一个经典难题，从七十年代起图像分割问题就吸引了很多研究人员 为之付出了巨大的努力。但到目前为止还不存在一个通用的方法，也不存在一个 判断分割是否成功的客观标准。本文对图像分割方法，特别是近几年来图像分割 领域中出现的新思路，新方法，或对原有方法的新的改进给出了一个比较全面的 综述。 针对CT图像的多阈值分割问题，我们对基于最大熵原则的多阈值分割方法做 了研究。最大熵原则是图像处理中阈值选择的最常用方法之一，很多研究者针对 最大熵原则作了各种研究，但一般回避了用最大熵原则进行多阈值选择时运算量 太大的问题。另外，H.D.Cheng等人将模糊性引入最大熵原则，提出了模糊最大 熵的概念，本文对他们提出的这种模糊最大熵原则做了修正，提出了我们自己的 模糊最大熵公式，并针对多阈值选择中计算量太大的问题，提出了解决这一问题 的ICM算法。 针对一般的医学图像序列的分割问题，本文介绍了一种结合live wire算法和 活动轮廓模型的医学图像序列的分割方法。我们通过把live wire算法和图像分割 中一般的区域增长方法结合来改进live wire算法，并用改进后的算法来对医学图 像序列中的单张或多张切片进行交互式的准确分割。然后计算机利用活动轮廓模 型来自动分割相邻的未分割切片。我们通过在活动轮廓模型的边缘点中引入记录 己分割物体边缘附近局部区域特征的灰度模型来把已分割切片中的物体与背景的 局部区域特征带入相邻的未分割切片中，并用由灰度模型定义的区域相似性代替活动轮廓模型中的外能来引导边缘轮廓收敛到物体的实际边缘。本文还介绍了一种基于live wire算法思想的简单的分割结果交互式修复方法。
英文摘要	Automated Fingerprint Identification System (AFIS) has widespread usage in practice. Fingerprint image enhancement and minutiae matching is two of the key problems in AFIS. In this dessertation, we proposed a fingerprint image enhancement algorithm based on orientation field. In addition, we proposed a minutia matching algorithm which modified Jain et al.'s algorithm. In our algorithm, a simpler alignment method is used. We introduced ridge information into the minutiae matching process in a simple but effective way. One of the advantages of doing so is we solved the problem of reference point pair selection with low computational cost. In addition, we used changeable sized bounding box to make our algorithm more robust to nonlinear deformation between fingerprint images. Experiments done on the FVC2000 databases with the FVC2000 performance evaluation method show that our improved algorithm is better than the original one. Image segmentation is one of the classical difficult problems. Many researchers had worked on this problem ever since 1970s. But no method that had good result for general images had been proposed and no impersonal criterion for deciding whether the segmentation is success had been generally accepted up to now. In this dessertation, we give a rather complete survey to the image segmentation methods, especially those new ideas, new methods and new improvement to the classical methods appeared in the literature of the last few years. The maximum entropy approach is one of the most important threshold selection methods in image processing. Many authors avoid the problem of computationally prohibitive when the maximum entropy criterion is applied to multi-level threshold selection. This dessertation proposed to deal with this problem using ICM(iterated conditional modes) algorithm. Experiments done to compare our ICM algorithm with the simulated annealing algorithm proposed by Cheng et al. fully disposed the effectiveness of ICM algorithm. The maximum entropy approach is one of the most important threshold selection method, H.D.Cheng et al. introduced fuzziness into maximum entropy approach and proposed the fuzzy maximum entropy criterion. In this dessertation, we do some modifications to their method, give our own formulation and propose to solve the problem of computationally prohibitive when the fuzzy entropy criterion is applied to multi-level threshold selection using ICM algorithm. Experiments comparing our ICM algorithm with the simulated annealing algorithm proposed by them fully disposed the effectiveness of our method. In this dessertation, we propose an algorithm for the semiautomatic segmentation of medical image series based on the combination of the live wire algorithm and the active contour model. We modify the traditional live wire algorithm by combining it with the region growing method and obtain accurate segmentation of one or more slices in a medical slice seri
语种	中文
其他标识符	657
内容类型	学位论文
源URL	[http://ir.ia.ac.cn/handle/173211/5721]
专题	毕业生_博士学位论文
推荐引用方式 GB/T 7714	罗希平. 生物信息处理:对自动指纹识别和医学图像分割的研究[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2001.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。