Mutual spectral residual approach for multifocus image fusion

Multifocus image fusion is the process of obtaining a single image from multiple partially focused images such that the newly formed image consists of the well-defined information extracted from each source image. This paper proposes the use of saliency of the source images based on Mutual Spectral Residual. Spectral Residual brings out the unique/salient features of an image in frequency domain. The idea of proposed mutual spectral residual is to emphasize the relative unique features of a source image with respect to the other source images. The relative unique features are utilized to form saliency maps for each source image. These saliency maps can clearly indicate the focused and defocused parts of an image. Based on the saliency maps obtained, the image fusion takes place in spatial domain. Visual inspection and quantitative evaluation of the fused images obtained by the proposed method, using different evaluation metrics, demonstrate its effectiveness over several existing image fusion methods.     

Analysis of distribution time of multiple files in a P2P network

Peer-to-Peer (P2P) file sharing attracts much attention due to its scalability and robustness. One important metric in measuring the performance of a P2P file sharing system is the amount of time required for all peers to get the files. We refer this time as the file distribution time. Researchers have proposed protocols to minimize the file distribution time under different situation. However, most works are based on the single-file scenario. On the other hand, there are studies showing that in a file sharing application, users may download multiple files at the same time. In this paper, we analyze the minimum time needed to distribute multiple files. We develop an explicit expression for the minimum amount of time needed to distribute multiple files in a heterogeneous P2P fluid model. Unlike the single-file scenario, we demonstrate that the theoretical lower bound in multi-file is not always achievable. With a comprehensive consideration of all the configurations, we elaborate how to partition the bandwidth capacities of both seeds and leechers for a particular file such that the finish time is optimal.     

Web日志在协同过滤推荐算法中的应用

协同过滤算法近年来在电子商务推荐系统中得到了广泛的应用,但该算法也存在数据稀疏性和缺乏个性化等问题,这些问题影响了推荐算法的效率和准确性。主要针对以上问题,提出引入Web日志分析的协同过滤算法,将用户对商品的隐性兴趣转化为显性兴趣,同时利用用户聚类等相关技术,不仅解决数据稀疏的问题也提高推荐的准确性。     

浅议学校档案管理的信息化建设

学校档案建设对于学校管理具有重要的作用。随着办公自动化的普及,档案管理工作也越来越多的运用网络技术进行现代化管理,现代化学校应当建立起现代化的档案管理体制,实现档案管理信息化、自动化。     

Single Sign—On（SSO）技术探讨

当为了支持商业进程而沿伸IT系统时,用户和系统管理员为了完成工作,面对着日益增长的复杂的各种接口（人机交互接口）。用户需要登录多个的系统,而系统管理员则需要以同样的方式,维护多个系统的用户账号。单点登陆系统能集统一各个系统的认证和授权,解决用户和管理员的困扰,并大大提高安全性。利用Kerberos思想和SSL技术,在HTTP协议基础上建立有效的认证和授权机制,提出可行的Web单点登陆平台解决方案。     

批量文件重命名VB程序设计

用VB6．0编程实现批量文件的重命名,新文件名由“固定部分”＋“顺序变化的序号”组成,扩展名不变,一次重命名文件数量最大可迭99999个。程序小巧,功能强大,界面友好,操作简单。     

浅谈PowerPoint在制作多媒体课件的应用技巧

当今信息社会,利用PowerPoint制作出与众不同的课件来协助我们教学,是经常利用多媒体辅助教学的教师们所感兴趣的问题。一般像简单的插入图片、音频、视频是大多数教师都能易如反掌的运用的。本文在掌握常用方法的基础上,对音频、视频、FLASH媒体文件的应用技巧进行了综合应用,主要说明如何运用一定的技巧,使课件质量得以提高,效果更加人性化。     

基于极坐标Log Gabor小波的纹理分析方法

提出一种基于极坐标Log Gabor小波的纹理分析方法,该方法突破了Gabor小波的带宽限制,仅采用较少的极坐标Log Gabor滤波器,即实现了准确地纹理分割.实验表明,在频谱覆盖相同的情况下,极坐标Log Gabor小波具有较Gabor小波更好的性能和更高的效率;此外,实验中针对多通道滤波方法对最小距离判别方法的改进,也取得了较好的效果,这为基于小波分解的分类问题提供了一个新的思路.     

可扩展的Web日志采集框架的设计与实现

以某网站所需的日志分析需求为背景,设计并实现了一种Web日志分析系统的前端日志采集框架.该日志采集方案没有采用传统的写日志文件然后轮训的方式进行日志收集,而是采用客户端/服务器的模式进行"自定义"格式的日志采集.该框架同样完全兼容传统的日志记录格式和记录方式.     

分布式大规模监控视频存储系统THNVR

通过分析网络视频监控应用中的存储需求特点,提出了一种采用低成本SATA硬盘的高性能存储解决方案。该方案使用定长文件,并将结构化与非结构化数据分别存储和检索,彻底克服了传统解决方案中磁盘碎片导致存储性能下降的问题。该方案还采用了自适应的缓冲技术,对磁盘访问进行了有效的规整和调度,充分利用了磁盘的写入带宽。此外,该方案还实现了容错、扩展等重要功能。性能分析显示,该方案在普通计算机和SATA磁盘上能够同时记录256路D1或512路CIF视频数据流。