【作者】 朱旭娟；

【导师】 沈兰荪； 李晓华；

【作者基本信息】 北京工业大学 ， 电路与系统， 2004， 硕士

【摘要】 摘 要近年来，信息网络的高速发展和多媒体业务的普及极大地改变了人类自身的生产、生活方式，人们对多媒体信息业务的需求促使新的多媒体业务的开发成为当前信息领域的热点。图像数据海量、无序等特点决定了大部分图像信息业务的构建都必须解决图像数据的高效压缩和有效检索两个关键问题。目前图像的高效压缩已经取得了长足发展，图像普遍以压缩格式存在，为了进一步加速图像信息的检索过程，基于压缩域的图像/视频检索技术的研究受到了人们的关注，压缩域基于内容的图像检索系统的开发也引起了广泛的重视。本论文通过对影响图像检索性能的多个关键技术的研究，基于 VB6.0 和数据库 SQL Server 2000 开发了一套压缩域基于内容的图像检索系统。主要工作包括：1．讨论了目前典型的几个原型系统，并从用户的角度出发，通过对目前图像检索系统基本框架结构的分析，设计了一个界面友好、能够对图像数据库进行有效管理且灵活、安全的图像检索系统；2．提出了一种用于提高图像检索精度的，基于自适应最优特征维的相关反馈算法，使检索趋势能够按照不同用户的意志自适应改变；3．讨论了多种用于提高图像检索响应速度的高维索引关键技术，并结合不同树的优点，提出了一种混合树算法，改进了图像的检索响应速度；4．针对图像上传和入库工作的复杂性，提出类模块式入库的方案，通过实验对其可行性和有效性进行了讨论。更多还原

【Abstract】 With the booming of digitized information and network, the patterns ofproduction and life of human beings have been changed dramatically due to the rapiddevelopment of information network and multimedia applications. The increasingneed for multimedia information has made the development of new network-basedmultimedia applications become one of the hottest subjects in the field ofinformation technology. Due to the enormous and unstructured multimedia data,solutions must be provided for their effective compression and efficient indexing inorder to realize all kinds of multimedia applications. Recently, effective compressiontechnique has substantially developed and images exist widely in compressed format.To make fast retrieval, indexing techniques of image/video data incompressed-domain have witnessed a booming interest. And people attachimportance to the research and development of content-based image retrieval systemin compressed-domain extensively. This paper discusses several key techniques which influence image retrievalperformance and develops a new content-based image retrieval system in compressed-domain based on VB6.0 and database SQL Server 2000.The major contents are: 1. Some current typical contented-based image retrieval systems are reviewed.Based on the structure analysis of these systems, this paper presents a flexible andsafe system with a friendly user interface and which can provide efficient imagedatabase management. 2. A novel relevance feedback algorithm using the optimal feature componentsadaptive extraction is put forward. Retrieval results can be changed with differentusers. 3. A new compound high-dimension indexing method is constructed by takingadvantage of some other high-dimension techniques. Simulation results show theefficiency of the method. 4. A model method according to the complexity of uploading and downloadingimages and features is proposed whose feasibility and efficiency are verified.更多还原