摘要： 基于语义树(Vocabulary tree)的图像检索方法是效果最好的方法之一，但目前存在的基于Vocabulary tree的方法都是建立在一种特征上的，当图像库比较大时很难达到理想的效果。基于此，提出一种多特征检索结果的融合框架Feature forest，根据各种特征的检索结果好坏动态确定对应特征树的权值。实验结果证明，相对于单种特征的特征树，该方法有一定的优越性。

关键词: 语义树, 特征融合, Feature forest框架, SURF特征, HOG特征

Abstract: Vocabulary tree-based method is one of the most effective image retrieval methods. However, for existing Vocabulary tree methods, the retrieval precision in large scale image database is not acceptable especially for image datasets with high variations. This paper proposes a novel tree fusion framework Feature forest, utilizing and dynamically fusing different kind of local visual descriptors to achieve a better retrieval performance. Experimental results show the effectiveness of the approach compared with single Vocabulary tree-based methods on different databases.

Key words: Vocabulary tree, feature fusion, Feature forest framework, SURF feature, HOG feature

中图分类号: 

• TP391.41