摘要
本文针对大型高维图象数据库检索,提出基于单特征进行高速、快捷的学习矢量量化算法,在独立的训练集(7537幅图象)和测试集(5000幅图象)上,通过实验对户内/户外,城市/风景分类,并在同条件下同贝叶斯分类器及支持向量机的结果比较。在计算复杂度和消耗时间上大大缩减,分类效果仍保持基本相似,且对有的分类问题可提高判别准确率。
An efficient method for fast-indexing image retrieval based on simple single feature is proposed. The algorithm of this method performs well on large high-dimension image databases. Good results have been attained from experiments for both Indoor/Outdoor and City/Landscape problems on two independent image databases separately. There are 7537 images in the trained set, while the tested set includes 5000 images. Compared with Bayesian Classifier and Support Vector Machine, Using this method, the executive speed can be accelerated greatly.
出处
《电路与系统学报》
CSCD
2002年第2期22-25,共4页
Journal of Circuits and Systems
基金
国家"九五"重点科技攻关资助项目(96-B02-01-05)