摘要
相关反馈是近年来基于内容图像检索的研究热点,本文将当前统计学习理论的最新成果——支持向量机应用于反馈过程,对用户给出的正负反馈图片信息进行学习,并在2035幅图片库上应用不同的特征组合进行了反馈学习过程的比较实验,探讨了图像特征对支持向量机性能的影响。实验表明,多类别的特征选取有助于检索性能的提高。
Relevance feedback is the hot issue in the research of co ntent-based image retrieval. This paper applies the newest theory about statistical learning at prese nt, support vector machine (SVM), in the process of feedback. It studies the information of positive and negati ve images which user provides. We compare the process of study using different features in a large database of 2 035 natural images and discuss about the effect on SVM's performance which image features cause. From the empirical results, we can conclude that multi-features in different classes for image can speed up performa nce of retrieval.
出处
《现代计算机》
2004年第4期14-16,共3页
Modern Computer
