基于目标轮廓形状矩阵傅氏描述子的特殊标志识别方法  

Method for special label retrieval based on shape matrix Fourier descriptor of object contour

作　　者：王新建[1] 罗光春[1] 秦科[1] 田玲[1] 彭凝多 赖云一[1] 

机构地区：[1]电子科技大学计算机科学与工程学院,成都611731

出　　处：《计算机应用研究》2015年第4期1231-1235,共5页Application Research of Computers

基　　金：中央高校基本科研业务费资助项目(ZYGX2013J071);四川省科技厅资助项目(2013JQ0005)

摘　　要：为了快速地从互联网上的海量图像中检索出含有某种特殊标志的图像,提出了一种基于形状矩阵傅氏描述子(shape matrix Fourier descriptor,SMFD)的图像标志检索算法。该算法通过对图像内容进行分割得到目标对象的边界信息,并在光栅系统中进行目标轮廓边界点统计获得形状矩阵,然后分析其周期性变化规律和特点,对形状矩阵按列展开为一维向量并进行傅里叶变换,取傅里叶变换系数中模值大于模值平均值的部分来构建特征向量,最后用欧氏距离进行图像间相似性度量。实验结果表明,SMFD具有尺度、旋转、平移不变性,与其他方法进行检索对比,提高了图像的查准率和查全率,可以有效地应用于实际项目。In order to quickly retrieve the images that contains some special label from massive images over the Internet,this paper proposed a method of shape image retrieval based on shape matrix Fourier descriptor. It extracted the corner information of the object by dividing the object contours,counted the corner points in raster coordination system to obtain the shape matrix,then analysed the periodic variation and its characteristics,expanded it by column as a vector and performs Fourier transform,constructed the shape of a matrix Fourier descriptors by taking the part where the Fourier transform modulus was greater than the average value,and finally measured the similarity among the images with Euclidean distance. Experimental results show that the SMFD is content to transfer,rotation and scale invariant,it is efficient and reliable compared to other methods with the images intercepted on the Internet. Therefore,it can be applied in the real project efficiently.

关 键 词：特殊标志 形状矩阵傅氏描述子 特征向量 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]