摘要
提出了一种基于形状矩阵的傅里叶描述子,该算法首先对植物叶片图像二值化,随后将图像的直角坐标转换为极坐标,并将极点移至图像质心,再利用极坐标光栅系统对图像重新采样形成形状矩阵,之后对矩阵进行离散傅里叶变换,利用低频系数描述图像的形状信息。实验结果显示该方法的检索性能明显优于Hu不变矩、等面积同心圆等其他5种算法,在对图像的抗噪性实验中也证明了该算法的高效性。
A Fourier descriptor based on the shape matrices is proposed in this paper.Firstly,the plant leaf images are converted to binary images,and then the rectangular coordinates of the image are converted to polar coordinates and the pole moves to the image centroid.Polar coordinate raster system re-samples images to form the shape matrices,followed by discrete Fourier transform,finally uses the low-frequency coefficients as image shape information.Experiments show that it is superior to other algorithms such as Hu's invariant moments,concentric circles of equal area and so on.The anti-noise experiment also shows efficiency of the algorithm.
引文
[1]闫雅楠,夏定元.结合边缘检测和区域分割的形状特征提取[J].电视技术,2007,31(3):12-15.
[2]魏怡,何一伟,倪海峰,等.形状描述法在图像检索中的应用综述[J].系统工程与电子技术,2009,31(7):1755-1761.
[3]LU G J,SAJJANHAR A.Region-based shape representation and simli-larity measure suitable for content-based image retrieval[J].MultimediaSystems,1999,7(2):165-174.
[4]GOSHTASB A.Description and discrimination of planar shape usingshape matrices[J].IEEE Trans.Pattern Anal.Mach.Intell.,1985,7(6):738-743.
[5]王斌.一种不变的基于傅立叶变换的区域形状描述子[J].电子学报,2012,40(1):84-88.
[6]HU M K.Visual pattern recognition by moment invariants[J].IRETrans.Information Theory,1962,11(2):179-187.
[7]卢蓉,范勇,陈念年.一种提取目标图像最小外接矩形的快速算法[J].计算机工程,2010,36(23):178-180.
[8]孙强强.基于内容的二值商标图像检索研究[D].扬州:扬州大学,2008.