商标图象库检索的多层次结构方法

基于商标图象库应用背景和具体特征，提出一种商标图象检索的多层次方法．图象预处理阶段采用人机交互方式下３种半自动方法提取商标图象各区域对象．检索首先采用字符串匹配方法和传统数据库查询技术，动态减少图象搜索范围；在图象区域规格化下，相继用最小约束矩和基于Ｈａｕｓｄｏｒｆ距离的相似性函数度量商标相似性程度，具有平移、旋转和尺度变换下的不变性和一定的抗噪声能力．实验结果表明，该方法有效和实用．

Hierarchical Structure Method for Trademark Database Retrieval

A hierarchical structure method for trademark database retrieval is proposed based on application background and idiograph features. During the image preprocessing stage three semiautomatic methods are used to extract image regions under man matching iteration. String matching algorithms and conventional database techniques are used to reduce image searching range. Smallest enclose rectangle and Hausdorff distance under region normalization are used to measure the similarity between two trademarks, and it is of invariance under transforms(translation, rotation and scale) and tolerance to noise. The experimental results show its efficiency.