基于Bayesian的相关反馈在医学图像检索中的应用  

Application of relevance feedback based on Bayesian theory in medical image retrieval

作　　者：张泉[1] 邰晓英[1] 

机构地区：[1]宁波大学信息科学与工程学院,浙江宁波315211

出　　处：《计算机工程与应用》2008年第17期158-161,共4页Computer Engineering and Applications

基　　金：国家自然科学基金重大项目(the Grand National Natural Science Foundation of China under Grant No.60472099)

摘　　要：前期的研究表明,代表纹理特征的灰度共生矩阵在胸片图像检索中发挥了比较好的检索效果,相关反馈能够很好地实现人机交互,有效地提高检索效率。进一步研究了两种移动查询点的反馈方法在提高检索效率的特点,首先基于分块的纹理进行检索,然后利用Rocchio方法和基于贝叶斯最小错误率的反馈方法进行多次反馈检索。实验证明,基于贝叶斯最小错误率理论的反馈方法能更好地缩小"语意鸿沟",提高检索效率。The preliminary research indicates that the gray level co-occurrence matrix display the better retrieval effect in the sternum image retrieval and relevance feedback realizing man machine interactive retrieval can effectively enhance the efficiency. In this paper two feedback methods with moving inquiry have further compared.First,retrieval with the piecemeal texture,then retrieval with Rocchio feedback method and based on minimal Bayesian error rate feedback method.Experiment proves that based on minimal Bayesian error rate feedback method can reduce ＂the meaning gap＂ , enhance the retrieval efficiency.

关 键 词：灰度共生矩阵 相关反馈 移动查询点 贝叶斯最小错误率 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]