基于泽尼克矩与边缘梯度共生矩的商标检索算法

Trademark retrieval based on Zernike moment and edge grodient co-occurrence moment

为提高商标的检索精度,定义基于泽尼克矩与边缘梯度共生矩的商标检索方案。引入中值滤波,对商标实施去噪处理;分别提取商标的泽尼克矩(Zernike moment,ZM)和边缘梯度共生矩(edge gradient co-occurrence matrix,EGCM);根据得到的ZM,通过Euclidean距离度量方法,计算查询商标与商标库的全局相异性及平均全局相异性;将平均全局相异性作为阈值,与全局相异性实施比较,剔除不相关商标;根据EGCM计算局部相异性,联合全局相异性,通过对二者实施加权得到总相异性,以输出检索结果。实验结果表明,与已有商标检索方案相比,所提方案具有更高的检索精度与鲁棒性。

To improve the accuracy of trademark retrieval,a trademark retrieval scheme based on Zernike moment and edge gradient co-occurrence matrix was defined.The median filter was introduced to denoise trademark.The Zernike moment and edge gradient co-occurrence matrix of the trademark were extracted respectively.According to the obtained ZM,the global and ave- rage global dissimilarities between the query trademark and the trademark database were calculated using Euclidean distance measurement method.The average global dissimilarity was taken as the threshold and compared with the global dissimilarity to eliminate the global unrelated trademarks.Local dissimilarity was calculated according to EGCM,and the total dissimilarity was obtained by combining with the global dissimilarities.The retrieval results were outputted using total dissimilarity.Experimental results show that compared with existing trademark retrieval schemes,the proposed scheme has higher retrieval accuracy and robustness.