商标图像检索中子图像特征融合准则研究

该文对商标图像检索中的子图像特征融合准则进行了研究，根据子图像可信度对子图像特征融合准则进行加权改进。用Hu不变矩进行基于子图像特征融合的商标图像检索实验，用PVR指数作为图像检索性能评价准则。实验表明，基于子图像特征融合的商标图像检索优于基于全局图像特征的检索，子图像特征融合加权准则优于未经加权的子图像特征融合准则，其中最小加权平均准则最优。

Research on Sub-image Feature Fusion Rules in Trademark Image Retrieval

This paper researches the sub image feature fusion rules in the trademark image retrieval, and improves the sub image feature fusion rules by weighting based on the sub image reliability. Experiments are performed using Hu invariant moments on the trademark image retrieval based on sub image feature fusion,and the PVR value is used as the image retrieval performance evaluation criterion. Experiments show that the trademark image retrieval performance based on sub image feature fusion outperforms the one based on global image features, and that the sub image feature fusion weighted rules outperform the unweighted ones, in which the minimum weighted average rule is optimal.