基于内容图像检索中的"语义鸿沟"问题

目的探讨目前CBIR系统中广泛存在的“语义鸿沟”问题。方法阐述了该问题的表现及产生的实质;从获取图像语义的角度出发,研究和分析了当前针对这一问题的一些处理方法以及存在的问题,并提出了初步的解决思路。结果现阶段在CBIR中应用多层次的相关反馈方法,能够较好地建立和修正图像低层特征与高层语义间的联系,有助于缩小图像检索过程中的“语义鸿沟”,并达到了一定程度语义检索的目的。结论实现真正意义上的基于语义的图像检索,将是解决“语义鸿沟”问题的最有效途径。

Review and research on "semantic gap" problem in the content based image retrieval

Aim To discuss the "semantic gap" problem which exists in the CBIR.Methods The form and origin of the problem are explored;from the point of acquiring the image semantics,the methods and their current lacks on solving the problem are studied and analyzed,and towards them some resolving strategies are presented elementarily.Results At present the applying multilayer relevance feedback methods in the CBIR could build and modify the association low-level feature and high-level semantic,which would be helpful to narrow the "semantic gap" in image retrieval and reach the purpose of semantic retrieval at some extent.Conclusion To achieve the real semantic-based image retrieval would be the most effective approach to settle the problem.