Abstract:The high incidence of brain disease，especially brain tumor,has increased significantly in recent years．It is becoming more and more concernful to discover knowledge through mining medical brain image to aiddoctors’diagnosis．Image mining is the important branch of data mining．It is more than just an extension of data mining to image domain but an interdisciplinary endeavor．Image clustering and similarity retrieval are two basilic parts of image mining．In this paper,we introduce a notion of image sequence similarity patterns(ISSP)for medical image database．ISSP refer to the longest similar and continUOUS sub．patterns hidden in two objects each of which contains an image sequence．These patterns are significant in medical images because the similarity for two medical images is not important，but rather,it is the similarity of objects each of which has an image sequence that is meaningful．We design the new algorithms with the guidance ofthe domain knowledge to discover the possible space occupying(PSO)in bain images and ISSP for similarity retrieval．The experimental results demonstrate that the results of similarity retrieval are meaningful and interesting to medical doctors．

Abstract:Knowledge acquisition is the bottleneck in developing expert system．It usually takes a long period to acquire disease knowledge using the traditional methods．Aiming at this problem，the paper presents the relationship between rough sets and rule-based rapeseed disease knowledge，namely the application of rough sets in knowledge acquisition from rapeseed disease expert．Then the exclusive rules，inclusive rules and disease images of rapeseed disease are built based on the RHINOS diagnosis model，and the definition of probability rule is put forward．At last，the paper presents the rule-based automated induction reasoning method，including exhaustive search，post-processing procedure，estimation for statistic test and the bootstrap and resampling methods．The results of experiment show that rough sets not only are a good framework for knowledge acquirement，but also Can accurately induct the rules of plant diseases．This method Can act as the assistant tool for development of diagnosis expert system，and has a extensive application in intelligent agriculture information systems．

Abstract:In this paper, the methods are investigate for online，frequent paRem mining of stream data,with the following contributions：(1) based on heuristic methodology and sample theory,step-by-step data stream mining method is used to estimate potential paRern set；(2)will find any length paRern not only single item pattern；(3)to find more appropriate length of each segment satisfying accuracy requirement,Hoeffding bound theory was introduced and revised to make it more suit for pattern mining；(4)a maintenance approach for estimating frequent patterns is developed for on．1ine analysis．Based on this design，estimation and maintenance algorithms are proposed for efficient analysis of data streams．This performance study compares the proposed algorithms and identifies the most accuracy-，memory-and time-efficient algorithms for stream data analysis．

Abstract:Software architecture has the potential to provide a foundation for dynamic software evolution．In this paper,based on a specific software architectural style，the dynamic evolut ion mechanism for distributed system are studied and put into practice．Firstly software architectural style named JB/HMB is described based on hierarchical message bus．Then the dynamic evolution types and processes，which are supported by JB/HMB，are dis cussed， Finally, ZIS，an implementation of hierarchical message bus，is proposed to support dynamic evolution of distributed system adhered to JB/HMB．

Abstract:Quality of services is a key to next generation Internet．More people focus on these problems，such as active queue management and random early detection．In this paper,we present an algorithm of adaptive RED，which can modify the max probability of packets dropping．We modify this parameter based on calculating the equilibrium point，so we can get the steady-state faster than some other methods based on self-adjustment．

Abstract:Reusing the components from the component repository is a way for software reuse．However,finding appropriate software components is often challenging．In order to help the users to find their components，the decision tree method is introduced to component repository to help inexperienced users’retrieving．Based on this standpoint，the process of retrieving component from component library is analyzed at first；secondly,the solutions about how to build the decision tree，how to retrieve components from the decision tree and how to save the decision rules are discussed；and then the method of decision tree based component retrieving is presented．Further more，a prototype decision tree based component retrieval system is introduced，and a preliminary experiment is presentedtoillustratethemethodinthe end ofthispaper．

Abstract:Set is a common data type in database system today．But there is no efficient index structure for set type data to support the queries relate to it．This paper presents a structure called SetUie．The stlxlcture is built based on the common prefix patterns in database．Unlike invert file，the sets with salne value are well organized．So the size of the data accessed by a query is smaller than that of invert file．This feature will cause the improvement of the selection operation’s performance．The experiments support this result．In this paper We also discuss several eptimizations approaches to Settrie．

Abstract:Usually,image zooming is an interpolation of the original image，achieving high quality zooming image as fast as possible．But it is well known that many traditional interpolation methods such as bilinear or bicubic often suffer from blurring edges or introduce zigzag errors．In this paper,a new adaptive image interpolation method based on edge，introducing rectangular interpolation and quincunx interpolation，putting forward the gradient weight function，motivated by the work of Nira Shezaf et al.，is presented．The experimental results show that the new method carl improve the quality of zoomed image effectively．At last，an evaluation is presented to assess the quality ofzoomed image．

Abstract:In order to support to the princir'le of least pfivUege effectively,considering the fimitatinns of traditional privilege mechanisms，a new Linux privilege mechanism called controlled privilege framewo (CPF) is proposed．CPF provides a fine-granularity partition of system privileges；improves the privilege computing mechanism of privileged process；and introduces the notation of privilege state for privilege control，refines the unit of pdvilege control farther．Based on CPF,fine-granularity and automatic privilege control can be performed totally transparent to all applications．The experimental results show that the threats of introsion are reduced and effective support to the prineiple ofleast privilege can be achieved．

Abstract:During the past two decades Model Checking based on Kripke Semantics Structure has proven its efficacy and powerful in circuit design，network protocol analysis，program verification and bug bunting．Recently there has been considerable research on Model Checking without OBDDs such as using SAT Solver or Bounded Model Checking (BMC). A Dynamic Kxipke Semantics Structure is introduced through allow the AP set changed．Based on this method．a new direct model checking algodthm is proposed．

Abstract:Nowadays data stream processing is becoming the new hot field of database research．Due to the volume and rapidness of data in stream，conventional techniques of query processing won’t be suitable any more. In such an environment a query is approximate．Histogram is commonly used to describe the distribution of data．This article presents a new algorithm ofmaintaining histogram under limited memory and guaranteed error, Experiments show that the algorithm is practical and efficient．

Abstract:In many scientific researches，people try tO discover the hidden relafionships among factors of some certain phenomena Most of these relationships can be expressed as functions．This paper analyzes the features of function expression and in．educes a function discovery method based on Gene Expression programming．The MEM method can deal with complex functions having n expressions(n>1)in difierent domains as well as those having only one uniform expression．The article also evaluates the complexity and performance of the method and shows that the MEM method has telarively low complexity due to the use of BDM algorithm and proves that the probabilityof successfulminingis 20％～80％，the averagetime elapsedislessthan 10s．

Abstract:For many KDD applications，such as data cleaning，detecting criminal activities in E-cormmerce，etc．finding the outlier call be more meaningful and interesting than finding the common cases．In the paper,we present a novel and efficient subspace Iocal outlier test algorithm：EDOLOIS．so as to avoid the computation-intensive distance computation．The algorithm takes full use of the character of subspace data processing and the initial LOF itself, thus it can not only reduce the computation dramaticaliy,but also gain the precise LOF of all objects in the subspaces．Both formal analysis and comprehensive performance evaluation show that the method is efficient to nnd all local outliers from high-dimensional categorical datasets in all subspaces．

Abstract:With XML being the standard for data encoding and exchange over Internet,how to manage XML data efficiently becomes a critical issue．An effective approach to improve the performance of XML management systems is to discover frequent XML query paRems and cache their results．Since each XML query call be modeled as a tree structore．the problem of discovering frequent query patterns can be reduced to frequent structure mining．However,mining frequem query patterns is much more complex than simple stroctnre mining since we have to consider the semantics of query patterns．In this paper,we present an approach to discover frequent XML query patterns efficiently．Compared with previous works，onr approach is strictly based on the semantics of XML queries，its mining results arc more precise，and Call be more effectively utilized by caching systerm.

Abstract:Web services Composition technology is gaining more and more attention．In this paper,the key technologies in Web Services Composition are surveyed，followed by brief introduction to each open problem．This paper also discusses some traditional problems，such as QoS，fault-tolerance and transaction，that are facing new challenges in Web services composition．

Abstract:In order to test Web Services，a testing technology based on Design-by-Contract is proposed in this paper．Usually,there are two problems when the users are using Web Services．Firstly,they cannot locate the occurring errors during testing precisely. Secondly,it is difficult to test whether Web Services meet their requirements．Design-by-Contract is an effective method to improve software reliability．This paper proposes a testing technology for Web Services based on preconditions and postconditions，then extends the syntax of WSDL and implements a prototype on the Microsoft．NET platform，which can solve the above problems better．

Abstract:This paper analyzes expandable and extended elements of Object-Oriented Framework and the relationship between them．For such elements,this paper specially discusses framework extension constraint that framework users should follow,and gives specifications of extension constraints using OCL(object constraint language)．

Abstract:The copyright protection mechanism emphasizes preventing the intellectual property right from being violated beforehand，which Can not satisfy the demand lying in law practice that the same source feature of software should be determined．In this paper feature of executable code is analyzed from static and dynamic state and a same source feature measuring method is put forward．The running result of demo system supports the method well．The ssme source feature measuring problem of executable code is resolved primarily in this paper．

Abstract:Scramble algorithms，which are also called pure-position permutation algorithms，are commonly used as an image ancryption，are investigated in this paper．Unfortunately, they are frail under known-text attack．According to the weakness of pure position permutation algorithm,We put forward an effective decryption algorithm for aIl pure-position permutation algorithms．First,a summary of the pure position permutation image encryption algorithms is given by introducing the concept ofergodic nieces．Secondly,by using probability theory and algebra knowledge，the declyption probabilit：of pure-position permutation algorithms is verified theoretically． Then，by defining operation system of fuzzy ergodic matrices，we improve a specific decryption algorithm．Finally some simulation results are shown.

Abstract:In this paper,all approach to generate reusable components from ELF object files is presented．Compared with other software component encapsulation technologies，building reusable software component directly from binary form ELF files is a straightforward process．Symbol tables end relocation information encapsulated in ELF files earl be automatically reprogrammed into a self-contained reusable entity,By augmenting the component with syntactic interface descriptions，mbignity caused by duplicate names in ELF files is eliminated．Binary code scanning is employed to enhance the safety of the component．Further,to prevent unintentional or hostile modifications．the component is signed with an MD5 fingerprint．

Abstract:The technique ofimage interpolation is usually required in image zooming．In this paper we make some changes to Hausdorff distance based on the theory of Hausdorff measurement．Then we use the changed Hausdorff distance to evaluate the interpolation algorithms．And also we compare it with other evaluation methods and we can find that Hausdorff distance has better results and more identical to the visinn impress．

Abstract:Slicing is a well-known reduction technique．Most of the research on slicing is code-based．There has been limited research on specification-based slicing and model-based slicing．EFSM is a very important specification model，but a practical EFSM model is often large and complex and is hard to understand and modify．EFSM—based slicing is very useful to test and analysis EFSM models．The dependence analysis based EFSM dependence graph is much more than a graph reachability problem，so that traverse algorithms based marking visited nodes carl not to be used in EFSM dependence graphs．In this paper,we discuss dependence analysis in EFSMs，and define a dependence transform function and a reverse dependence transfer function to describe dependence transform formally．Based the analysis and defmition，a backward slicing algorithm is given to find all transitions which may affect an interesting transition．

Abstract:An image-based rendering method that can successfulty support the representation of 3-D surface details and view motion pasallax is presented．Expensive task of hole-filling is finished beforehand in the preprocessing stage, only tcxtLIre-mapping hardware is employed in the rendering stage for acceleration．For correct representation，each pixel in a parallel projection image with depth is decomposed into several consecutive layers so that a bar texture is generated for that pixel．Since all the bar textures as a whole constitute a continuous 3-D approximation to the scene represented in the original depth image，novel views can be obtained by mapping these bar textures in a from-back-to-fiont order．Experimental results show the new approach can produce high visual qualny and nlns faster than many existing image-based rendering techniques with only moderate storage overhead．

Abstract:A novel scheme of mesh editing is proposed based on geometry signal spectrum analysis，which effectively combines mesh deformation in space domain and the geometry signal processing techniques in frequency domain．For intuitive interaction purpose，a geometry proxy of the original mesh is constructed，the consistent parameterization for both mesh and proxy are established．With signal spectrum analysis and conveying the deformation information from the geometry agent to the original mesh，we finally achieve the goal of mesh editing．The crucial innovation is the optimization of parameterizational mesh feature alignment scheme based on the radial basis function method．Additionally,an adaptive sampling approach is also devised for better sampling original model geometry signal．As extensions of spectrum editing scheme demonstrated in this paper,the editing principle is applied to mesh morphing and geometry details transferring with satisfying results．

Abstract:Space deformation is an important technique for shape modification and flexible animation generation in geometric modeling and computer animation．A novel multiresolution space deformation method is proposed in the paper which is controlled by subdivision surface of arbitrary topology．It supplies a solution for the complex shape edit while method is fast enough to achieve realtime interaction．Several subdivision schemes are adopted and compared in the method．Experimental cdteria are given on how to determine appropriate subdivision scheme and subdivision depth．To achieve realtime interaction，parameterization is accelerated through space-subdivision and deformation is evaluated locally．All of above algorithms are integrated into a space deformation prototype system．The system has the functions of multiresolution space deformation and abundant control mesh generation methods．

Abstract:In extracting the skeleton from the 3D complex model，the existing methods may produce errors that some branches are away from the center positions and some branches are redundant．As for this，a rectification method is proposed in this paper to correct these errors．After a coarse skeleton is extracted based on the distance field of the 3D model，the method investigates every joint to see whether it is away from its local center position． The investigation work by omitting radial rays evenly with thejoint as the starting point to intersect the model’s surfaces and studying the length distribution of the segments from the joint to the intersection points along the rays.For the ioints away from the center positions，they are moved to the center and their related branches are also modified．Thus，the skeleton can be rectified automatically．Experimental results show the new method can efficiently rectify the skeleton and run fast，and generally the time for rectification takes up less than 2％ of the time for extracting the skeleton．

Abstract:An anisotropic dvnoising algorithm for point-sampled models is proposed in this paper．Point-sampled models obtained by 3D scanning devices inevitably contain some undesirable noises．Aiming at quickly removing the isolated noises and preserving the appearance of geometrical sharp features，the local spatial geometry and range components including normal and curvature information arc considerably taken into account．and the noises are anisotropieally diffused by applying Gaussian kernel function to compute influence weights of neighbors and moving vertex along normal direction．The method is compared with previous algorithms It is proved that it is simple and efficient．

Abstract:We present a hybrid sort-first and sort-last parallel rendering system based on．retained-mode，called In-the-Core．which can render extreme large modeis on PC clusters．Traditional parallel systems have to issue all primitive-instructions when draw a new flame，or to copy whole scene on each node of cluster in advance．The approach can utilize full main memory capacity of cluster and lower the traffic of network．The main method is to partition a large model into smaller structures of elementary particle and distribmes them into the main memory of rendering nodes；meanwhile the load-balance is kept by view-depend partition algorithm with overhead feedback．Experiments on a cluster of 8 nodes show that the In-the-Core system can render models about 30～40M triangles at interactive frame-rate. It shows that a cluster of inexpensive PCs is all attractive alternative to those high-end graphics workstations.

Abstract:A fairing algorithm with centroid constraints ofneighborhood is presented for point-sampled geometry．By optimizing a local function based on the centroid constraints of neighborhood, local smoothing is achieved without surface reconstruction．This method overcomes the problems of extreme shrinkage and extreme convergence arisen from Laplacian smoofaing operator so that it achieves little distortion．The experimental results verify that it is stable，fast and easy-to-use．

Abstract:Study the interval implicitization of rational B-spline curves，that is，finding a lower degree interval implicit B-spline curve which bounds a given B-spline curve such that the width of the interval implict B-spline curve is as small as possible．The problem is solved in two steps：finding an approximate implicit B-spline curve and finding the bounding implicit curves by solving some linear problems．Some examples are provided to demonstrate the algorithm．

Abstract:This paper extends the classical 4-point interpolating subdivision scheme，and brings forward a new 4-point subdivision scheme with threc parameters for curve design，which cad be confrolled by choosing these three parameters appropriately．The sufficient conditions of the uniform convergence proporty and eonfinuity properties ofthe subdivision scheme are proved.

Abstract:A real-time globa relighting algorithm is proposed in this paper．The scene illumination is divided into three parts：direct，indirect and ambient components．After calculating the reflectance of objcots from reconstructed genmetry and pre-sampled images，indirect and ambient components of each sampled image are fitted with low-order basis functions To relight the soofle under new illumination conditions or viewpoints，the recovered reflectance is applied to the illumination model to get direct lighting．while indkect and ambient ones are obtained from the basis functions．In addition surface derails are recovered by the variations of the micro-property and orientations on the sulfaee computed from sampled images．The results show that this algorithm call reconstruct global illumination with clear shadow and surface detail vividly in real-time．

Abstract:Based on the relationship between rational Bemstein basis and Bemstein-like trigonometric basis，and the properties of normalized B-basis of the space of m-order trigonometric polynomial，a method is obtained by which the basis of the space can be transformed into rational Bernstein basis．An efficient algorithm is developed to obmin the rational representation for segment of a class of transcendental curves which are in the space．

Abstract:An implicit Curve reconstruction method is proposed which represents the curve with all algebraic tensor-product B-spline，and minimizes the tension ofthe B-spline and the approximate geometric distance between the curve and the point set．The method is dynamic and self-adaptive based on trust-region algorithm in optimization theory．The specification of the initial shape is priceless，and the high-quality reconstruction corvc is obtained in a robust way．Some examples are implemented．

Abstract:A method of generating cubic spline curves with local shape parameters is presented in this paper．The given method takes the Hermite interpolation curvcs and the cubic non-uniform B-spline curves as the special cases and unifies the representation of the curves interpolating the control polygon and the cubic non-uniform B-spline curves approximating the control polygon．A shape parameter only influences two curve segments，and the expression of the curves retains the simple construction of the expression of the cubic Bezier curves．We can adjust the shape of the curves locally by changing the values of the shape parameters or adjusting the Bezier control points．Based on the given spline curves，the bicubic spline surfaces with local shape parameters are given．

Abstract:Realistic simulation of natural scene has been one of the most challenging tasks in compute graphics．And it is even more difficult to real-time simulate snowing scene because the complex modeling of wind field and the interaction between wind and snow are involved．In this paper,based on the fully consideration of the physical characteristics of wind and snow,a three-dimensional wind field is constructed by discretizing the classic Boltzmann equation．According to the interactive laws between wind and snow,the changing rules of snow’s falling，deposition and erosion are then established．Finally,by adapting a series of simplification and accelerating rendering techniques，realistic wind-driven snow scenes under different speeds of wind with different snow falling amount are rendered in real time．