摘要：

在基于内容的图像检索和分类系统中，图像的底层特征和高层语义之间存在着语义鸿沟，有效减小语义鸿沟是一个需要广泛研究的问题。为此，提出一种基于特征互补率矩阵的图像分类方法，该方法通过计算视觉特征互补率矩阵进而指导融合特征集的选择，利用测度学习算法得到一个合适的距离测度以反映图像高层语义的相似度。实验结果表明，该方法能有效提高图像分类精度。

关键词: 距离测度学习, 特征互补率矩阵, 特征融合, 图像分类

Abstract:

In content-based image retrieval and classification system, there is a deep semantic gap between low-level features and high-level concepts of image. The topic on how to bridge the semantic gap effectively needs to explore widely. This paper proposes an image classification method based on feature complement ratio matrix. A visual feature complement ratio matrix is computed, based on which the fusion feature set can be selected effectually. A proper distance metric, which reflects the semantic similarity between images, is obtained by using a distance metric learning algorithm. Experimental result shows that the method can improve the accuracy of image classification effectively.

Key words: distance metric learning, feature complement ratio matrix, feature fusion, image classification

中图分类号: 

• TP391.41