一种基于SVM的相关反馈图像检索算法

SVM-based Relevance Feedback Algorithm in Image Retrieval

相关反馈技术是近年来图像检索中的研究热点,本文以MPEG-7的边缘直方图作为图像特征,以支持向量机(SVM)为分类器,提出一种新的相关反馈算法。在每次反馈中对用户标记的相关样本进行学习,用历次返回的结果更新训练样本集,建立SVM分类器模型,并根据模型进行检索。本文还对不同核函数的SVM进行了对比,得出RBF核函数的SVM有较高的检索精度。使用由10000幅图像组成的图像库进行实验,结果表明,该算法可有效地检索出更多的相关图像,并且在有限训练样本情况下具有良好的泛化能力。

Relevance feedback technique has been an important approach in image retrieval. A novel relevance feedback algorithm is presented based on Support Vector Machine (SVM) in content-based image retrieval system, using MPEG-7 edge histogram feather. During the retrieval process, users can mark positive sample images similar to the query image. Then, the algorithm constructs a SVM classifier, which uses the updated training samples. The performances of SVMs with different kernel functions have been tested and compared. The RBF function has higher precision of retrieval than others. Experiments were carried out on a database of 10000 images. It shows that more images relevant to the query can be found efficiently by the retrieval procesS. It also shows the generalization ability of SVM under the conditions of limited training samples.