摘要
现有的图像检索系统多是针对底层特征的系统,而人类往往习惯于在语义级别进行相似性判别。如何跨越底层特征和高层语义之间的"鸿沟",成为基于内容检索的研究重点。本文提出一种利用SVM提取图像的高层特征,然后对图像进行语义级别的分类。实验结果表明,该方法在一定程度上跨越"语义鸿沟"。
image retrieval systems are mostly based on low-features,but people are accustomed to judging on high-semantic level.It becomes more important that overcome the considerable gap between them.This paper uses SVM to get the high-semantic concepts and classify the testing images.The results indicate that the SVM can span this gap to some extent.
出处
《微计算机信息》
2010年第1期115-116,156,共3页
Control & Automation
