基于内容的图像归档与查询中知识方案的研究

Research of Knowledge Scheme in Content- based Image Archiving and Query

在对基于图像内容的查询研究中认识到这是一个需要结合图像特征量提取和模式识别综合作用的过程。通过将知识库的求解分为两个层次:底层部分的特征量分类和在此基础上的上层符号系统,提出了一种人工神经网络连接系统和符号推理系统结合的基于内容图像查询算法中的知识方案。力求将问题中的矛盾分离,在每个阶段中着力于解决主要矛盾而采用相应的知识系统。该方案在一定程度上融合了两者的优越性,克服了各自的缺陷,在多特征量融合上也起到了一定的作用。最后给出了具体医学图像数据库应用系统中特征提取和知识系统的整体方案,并对试验结果进行了讨论,验证了上述方案的有效性和可行性。

During the research of Content- Based Images Query (CBIQ),we realize that it is a process in need of a combination of image feature extraction and pattern recognition.In this paper,a knowledge scheme combined ANN and symbol ratiocination system is presented and used in CBIQ.And it divides the solution of knowledge database into two levels.One is the feature classification understratum,and the other is symbol system based on it.So the main problem in each stage is separated and applied corresponding knowledge system. This scheme merges the advantages of the two knowledge systems and takes effects in multi feature syncretic.At last,an integer scheme applied in medicine image database is presented and the result of experiment is discussed.And it proves the validity and feasibility of the scheme.