在算法上

This thesis discusses the realization of the picture processing function of later phase in the process of PACS development. Basing on traditional digital image processing and according to the characters of medical images, we design the Mark Boff Chain edge detection algorithm basing on the distributing model of image grade scene, and has got good effect in the practical experiments; as to the system development, we design customizable DICOM standard analyzer, and use the trigger mechanism Callback to improve the analysis speed ,and adopt the system structure basing on the standard interface to reduce the modules coupling degree, and strengthen the stability of the system.

本文在PACS系统开发过程中，对其影像后期处理功能的实现进行了探索，在传统数字图像处理的基础上，根据医学图像自身所具有特点，设计出了基于图像梯度场分布模型的马尔可夫链边缘检测算法，并在实际测试中取得较好效果；在系统开发方面，设计出了可定制化的DICOM标准解析器，应用CallBack触发机制来提高解析速度，采用基于标准接口的系统架构体系降低了组件耦合度，增强了系统的稳定性。

This thesis is divided into six parts. The first chapter is preface, the current status of research in the inverse problems for parabolic partial differential equations is reported; the second chapter is "regularization methods for numerical differentiation and their applications ", in this chapter we investigate many regularization methods from a viewpoint of regularization theory and algorithm, some applications in the inverse problems for parabolic partial differential equations are given; the third chapter is "spectral regularization methods". Based on Fourier analysis, within the framework of regularization theory, we apply the spectral methods to some ill-posed problems. Many numerical experiments are done in order to show the validity of the methods; the fourth chapter is devoted to wavelet dual least squares method and a revised wavelet method; in the fifth chapter,we combine finite difference method with method of lines and apply it to the backward heat conduction problem in time; in the sixth chapter "identification problems for unknown source ", the essence and the degree of two problems related to source identification are pointed out, at the same time, some numerical methods are reported.

本文分为六个部分，第一章前言简要分析了国内外抛物型偏微分方程反问题的研究现状；第二章数值微分的正则化及其应用从正则化理论和算法的角度出发，考察了许多正则化方法，还给出了数值微分在抛物型偏微分方程反问题的一些应用；第三章谱正则化方法是在Fourier分析的基础上，在一般正则化理论的框架下，给出了这种方法在各种不适定问题中的应用，数值实验表明谱方法是有效的；第四章研究了小波对偶最小二乘方法和改进的小波方法；第五章主要研究了有限差分方法结合线方法在时间反向热传导问题中的应用；第六章是未知源识别问题，主要指出了两类未知源问题的不适定程度和不适定本质，同时报告了一些数值方法。

The evolutionary relationships were investgated by comparing the results from molecular data alone or molecular data combined morphological characters using different analysis methods: maximum parsimony by PAUP(superscript *) 4.0 and Bayesian inference by MrBayes 3.0B4. Base composition and ti/tv of 28S rDNA D2 regions were analyzed by PAUP(superscript *) 4.0. The results showed that GC content of the 28S rDNA D2 gene sequences of subfamily Euphorinae ranges from 40.00% to 49.25%. Transversion frequency of sequence variable sites among species of Euphorinae was higher than that of transition. The subfamily Euphorinae was proved to be a paraphyletic group with close relationship to subfamilies Neoneurinae and Cenocoelinae; only two tribes, Meterorini and Microctonini, were proved to be monophyletic groups; Centistini, Cosmophorini, Euphorini and Dinocampini were paraphyletic groups; the basal position of tribe Meterorini within Euphorinae was partially supported by our phylogenetic trees, and tribe Microctonini was proved to be a group at the top of the trees.

结果表明：优茧蜂亚科的28S rDNA D2基因序列片段的GC%含量在40.00%~49.25%之间变动，而对于碱基替代情况来讲，优茧蜂亚科各个成员间序列变异位点上颠换大于转换；不同的分析和算法所产生的系统发育树都表明目前根据形态定义出的优茧蜂亚科Euphorinae不是一个单系群，而是一个与蚁茧蜂亚科Neoneurinae和高腹茧蜂亚科Cenocoelinae混杂在一起的并系群；在优茧蜂亚科内部，悬茧蜂族Meterorini和食甲茧蜂族Microctonini（排除猎户茧蜂属Orionis）为单系群，而宽鞘茧蜂族Centistini、大颚茧蜂族Cosmophorini、优茧蜂族Euphorini、瓢虫茧蜂族Dinocampini为并系群；悬茧蜂族Meterorini在优茧蜂亚科Euphorinae内位于基部位置的观点得到部分的支持，同时食甲茧蜂族Microctonini被判定为相对进化的类群。

The synthesis scheme we developed can deal with changes in visibility, produce perspectively correct views and handle the distortion around the circumferences in the synthesized views. In our scheme, the area without corresponding points was first determined by using Fourier transformation in order to deal with distortion problem, within the mutual overlapped area the z-buffer values of matched key points were computed so that the visibility in novel views can be handle correctly, lastly the common points in new views are rendered through epipolar geometry constraint.

算法的步骤为：利用傅立叶变换方法确定相邻源图像无对应点的区域；然后，利用尺度空间和相位一致原理在共同区域提取及匹配边缘上的关键点，计算其深度值，并在新图像中由该值判别虚拟图像中的可见性；再由关键点恢复极线约束，通过极线间的位置和颜色插值生成虚拟图像中有对应区域的一般点；在确定极线位于无对应点区域段后，利用加权平均或放大插值的方法实现了源图像无对应点区域在虚拟图像中的自然过渡。

We mainly discuss three types of inpainting models: the Bayesian model, geometric models and PDE models with emphasis on investigating TV inpainting, elastica inpainting and Mumford- Shah inpainting of PDE models. After studying the level set method and its applications to image processing, we concentrate on discussing the inpainting model...

在介绍水平集法及其在图像处理方面的应用之后，进一步深入地讨论了一类图像修复的变分模型—Mumford-Shah-Euler模型，在此基础上利用水平集法，提出MSE模型的数值逼近的算法格式，并对一些具体的图像修复问题开展数值模拟和修复，深入分析了MSE模型及水平集法在图像修复中的优缺点，这些结果在理论和实际应用中都很有意义。

By using the kernel function, the input space is mapped to a high dimension feature space where the data are expected to be more separable, the initial centroids in the feature space are selected by applying the KRA algorithm, and the large and small clusters are partitioned and the outliers can be split from the large clusters iteratively after the kk-means clustering. As a result, the audit data can be clustered better. Secondly, the closed sequential patterns mining algorithm CloSpan is improved according to the restrictions that is composed of the axis properties and reference properties.

该方法通过核函数把数据样本空间映射到一个高维的特征空间，使数据在新的空间中具有更好的可分离性；在特征空间采用KRA算法选取初始聚类中心，然后在核k-means聚类的基础上，划分出大簇小簇并在大簇中分离出异类再次进行核聚类，从而不断地优化聚类结果。

Gas column (2) gaseous oil column (3) gaseous liquid column from pump to the tube end (4) gaseous column from tube end to the formation According to the flow pattern, this paper utilize the methods given by those papers .

从管柱结构上，将井筒划分为四个部分：(1)气柱段(2)沉没度(3)泵口至管鞋段(4)管鞋至油层中部在流态分布上，管鞋以上，尊重人前人的算法。

On the basis of taking full advantage of multi scale wavelet analysis,the proposed algorithm improves the contrast of the approximations of the wavelet transform through two lifting steps,and shrinkages the coefficients of wavelet transform on the diferent degree and then enhances the coefficients.

在多尺度分析基础上，该算法对小渡分解得到的低频子带图像采用两步提升法进行对比度增强处理，而对小波分解得到的不同方向上的小波系数进行不同程度的去噪并增强。

Based on the kinematic and dynamic analyses of the gear honing model where a worm-shaped hone is used, a theoretica.l study on the mechanical mechanism of the process is performed.. Looked upon.from, its principle of computation, the analysis of gearing is transformed into furthermore a problem of non-linear programming.

本文通过蜗杆珩轮珩齿模型的啮合运动、动力分析，探讨了珩齿的力学机理，并从计算原理上把啮合运动分析问题转化为非线性规划问题，在此基础上提出了计算机辅助珩齿机理分析的算法。

From the point of 3D scene imaging, comparing to the primary arithmetic, the result obtained from the improved arithmetic can preserve higher similarity with original models and has better display quality.

从三维成像效果上看，与普通二次型度量简化方法相比，改进算法得到的结果在外观显示上与原始模型能保持更高的相似性，有较好的显示质量。

The split between the two groups can hardly be papered over.

这两个团体间的分歧难以掩饰。

This approach not only encourages a greater number of responses, but minimizes the likelihood of stale groupthink.

这种做法不仅鼓励了更多的反应，而且减少跟风的可能性。