【作者】 蔡念； 张国宏； 楼朋旭； 戴青云；

【Author】 CAI Nian,ZHANG Guo-Hong,LOU Peng-Xu,DAI Qing-Yun (School of Information Engineering,Guangdong University of Technology,Guangzhou 510006,China)

【机构】 广东工业大学信息工程学院；

【摘要】 提出一种多特征融合的外观设计专利图像检索方法。形状特征融合了几何结构特征与边界方向直方图特征来描述,通过整体结构和轮廓特征的融合全面表达了物体的形状;纹理特征融合了Gabor小波提取的特征和本文提出的分块统计特征;最后采用加权法融合形状特征和纹理特征,实现了外观设计专利图像检索。实验结果表明,本文方法对外观设计专利图像检索有良好的效果。更多还原

【Abstract】 An image retrieval method for design patent is proposed based on multi-feature fusion.Structure and edge direction histogram describe shape features of design patent images.Texture features are obtained by Garbor wavelet and statistical characteristics of image blocks.Image retrieval for design patents is accomplished with shape features and texture features integrated with a weighted method.The experimental results indicate that the proposed method is feasible for design patent image retrieval.更多还原