摘要
相关反馈技术是近年来图像检索中的重要研究方向,它有效地缩短了用户高层语义和图像底层视觉特征的差距,大大提高了系统的检索精度。文中从机器学习的角度出发,提出了一种基于RBFN的相关反馈算法。同时,为了方便用户对检索结果的标记,将模糊逻辑引入到图像检索中。即:用户对检索结果标记为相关图像、模糊相关图像和不相关图像,利用这些反馈信息动态地建立RBFN的结构,并进行检索,这个过程反复进行直到用户得到满意的结果。实验表明,这种方法在图像检索中具有更好的性能和更强的推广能力。
Recently, the relevance feedback technique has been one of the important research facts in CBIR. Because it has greatly reduced the gap between the high level notion and low level visual features, the retrieval results are better. In this paper, proposed a relevance feedback approach using a network of radial basis functions in the view of machine learning. Meanwhile, integrated the conception of fuzzy logic for user's convenience on labeling the retrieval results. That is, the results are classified into relevant images, fuzzy relevant images and non - relevant images by labeling. Then make use of this information to dynamically construct RBFN, and retrieval images. The process is done repeatedly until the user is satisfied. The experimental result shows that the algorithm has better performance and generalization ability and is able to fulfill the user's requirement.
出处
《计算机技术与发展》
2007年第9期31-34,共4页
Computer Technology and Development
基金
上海海事大学基金资助项目(2005079)