###
DOI:
Journal of Software:2001.12(10):1479-1485

基于神经网络自学习的图像检索方法
张磊,林福宗,张钹
(清华大学计算机科学与技术系,北京 100084)
A Neural Network Based Self-Learning Algorithm of Image Retrieval
ZHANG Lei,LIN Fu zong,ZHANG Bo
()
Abstract
Chart / table
Reference
Similar Articles
Article :Browse 2795   Download 3052
Received:June 26, 2000    Revised:May 25, 2001
> 中文摘要: 相关反馈技术是近年来图像检索中较为活跃的研究方法之一.提出了一种基于神经网络自学习的图像检索方法,即在检索阶段利用人-机交互技术选出与检索图像相似的正例样本,然后构造出前向神经网络,进行自学习,以逐步达到提高查询效果的目的.神经网络的构造过程即是学习的过程,而且可以不断地学习.使用由9918幅图像组成的图像库进行实验,结果表明,该方法有助于用户表达查询意图和语义概念,可以通过交互式检索逐步求精地查找出更多、更准确的图像,并且具有较强的鲁棒性,可以结合各种特征表示和相似性匹配方法,交互地提高检索性能.
Abstract:In recent years, relevance feedback technique has become an active research method in image retrieval . A self-learning algorithm of image retrieval using forward propagation neural network is proposed in this paper. During the interactive retrieval process, users can mark positive images similar to the query image. Then the algorithm constructs a forward neural network and retrieves again based on the learned neural network. The experimental result over 9 918 images shows that the proposed approach greaty reduces the user's effort of composing a query and representing a concept. During the interactive learning and retrieval process, more and more correct images can be found in the anterior result. This approch is robust to various kinds of feature representation and simiarity distance formulas.
文章编号:     中图分类号:    文献标志码:
基金项目:国家重点基础研究发展规划973资助项目(G1998030509);国家自然科学基金资助项目(69823001);国家教育部博士学科点专项基金资助项目(98000335) 国家重点基础研究发展规划973资助项目(G1998030509);国家自然科学基金资助项目(69823001);国家教育部博士学科点专项基金资助项目(98000335)
Foundation items:
Reference text:

张磊,林福宗,张钹.基于神经网络自学习的图像检索方法.软件学报,2001,12(10):1479-1485

ZHANG Lei,LIN Fu zong,ZHANG Bo.A Neural Network Based Self-Learning Algorithm of Image Retrieval.Journal of Software,2001,12(10):1479-1485