基于贝叶斯理论的图像标注和检索  

Image Annotation and Retrieval Based on Bayesian Theory

作　　者：张元清[1] 包骏杰[1] 况夯[1] 钟洪[2] 

机构地区：[1]重庆教育学院,重庆400067 [2]中南大学信息工程学院,长沙410075

出　　处：《计算机科学》2008年第8期229-231,共3页Computer Science

摘　　要：图像自动语义标注是基于内容图像检索中很重要且很有挑战性的工作。提出用语义约束的聚类方法对分割后的图像区域进行聚类,在图像标注阶段,使用贪心选择连接(GSJ)算法找出聚类区域的独立子集,然后使用贝叶斯理论进行语义标注。对图像进行标注以后,使用标注的关键字进行检索。在一个包含500幅图像的图像库进行实验,结果表明,提出的方法具有较好的检索性能。Automatic semantic annotation of image is a crucial and highly challenging work in content-based image retrieval. A method of constrained clustering is proposed to cluster the segmented region, during the annotation stage, a greedy selection and joining method is used to find out the independent region cluster subsets. Then,a method of Bayesian is used to image annotation. After the image annotation,we can use the annotation keywords to image retrieval; the system has been implemented and tested on an image database of about 500 images. The experiment results show the effectiveness of the proposed approach.

关 键 词：图像分割 贝叶斯理论 约束聚类 图像标注 图像检索 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP312[自动化与计算机技术—计算机科学与技术]