Content-based image retrieval using shape similarity

Abstract

状特征由于其简单、直观等优点在基于内容的图象检索技术研究中非常重要。本文在简要介绍了基于内容的检索技术（Content－basedRetrieval，CBR）的基础上，从不变特征提取。相似性度量、检索方式等方面，详细探讨了利用形状相似性进行图象CBR的技术关键。实验结果表明，不变矩和相对面积等特征相结合可以有效地进行形状检测。相信本文的工作，将对构建某些特定应用的系统具有很好的参考价值。 Shape feathes is very essential for content - based image retrieval in terms of smplicity and intuition. In this paper, content - based retrieval (CBR) techniques are introduced firstly; based on which, methods for image CBR using shape similarities are discussed in detail including invariant feature extraction, similarity matching and retrieval mode. The satisfying experimental results showed that the comination of invariant moment and relative area is very efficient and effective for shape retrieval. It is hoped that the investigation will benefit the construction of domain specific applications for reference.