摘要： 利用整体结构特征和局部纹理特征的优势，采用两级分类器对场景图像进行分类。第1级分类器利用全局结构信息得到候选类别，并通过分类结果判定相似类别对；第2级分类器则利用局部纹理信息区分相似类别，采用分类器的级联综合利用场景图像的整体结构信息和局部纹理信息。实验结果表明，该方法能够做到不同场景类别鲁棒分类，有效区分相似场景类别，提高场景图像的分类准确率。

关键词: 结构特征, 纹理特征, 场景图像分类

Abstract: This paper proposes a scene image categorization method based on structure and texture fusion. It adopts a two-layer classifier. The first classifier classifies all the scene images based on structure features, and similar categories can also be computed based on the results. The second classifier only classifies similar images and both of results are combined to predict the scene image category. Experimental result shows that categorization accuracy can be improved based on the method.

Key words: structure feature, texture feature, scene image categorization

中图分类号: 

• TP393.41