摘要
文中针对二值商标图像,提出一种基于方向特征的商标图像检索方法。用边界方向直方图表示目标的边界方向特征,图像的相似性度量采用直方图求交算法;用区域方向直方图表示目标的区域方向特征,图像的相似性程度用直方图的相关距离来度量。这两者结合起来得到的方向特征可以同时反映图像的整体形状和轮廓形状。实验表明,方向特征能够有效地描述图像的形状及空间分布信息,取得了令人满意的检索结果。
This paper proposes an images retrieval method based on direction feature for binary trademark images,it use edge direction histogram to describe the direction feature of the boundary of object,the similarity is measured by histogram intersection algorithm,and it use region direction histogram to describe the global direction feature of image,histogram correlation distance is used to compare the similarity. By integrating the two different shape object shape attributes,the direction feature can reflect both the global shape,and the silhouette shape of the object. Experiments show that this method can capture the shape and spatial information of image effectively,and the retrieval result can fit with human visual feelings
出处
《计算机应用》
CSCD
北大核心
2003年第7期77-79,共3页
journal of Computer Applications
关键词
商标图像检索
方向特征
方向信号
区域方向直方图
边界方向直方图
trademark image retrieval
direction feature
direction signal
region direction histogram
edge direction histogram
