椭圆检测算法的比较

椭圆检测在图像识别与计算机视觉领域中一直占有重要的位置,而边缘检测是图像特征提取(包括椭圆检测)的首要条件.介绍了常用的几种边缘检测算法(如Sobel算子、Canny算子、Zernike矩算子等),分析了这些算法的异同与优劣;讨论了几种基于Hough变换及其改进方法的椭圆检测算法(包括利用椭圆极点与极弦性质的提取算法、利用三点确定椭圆的提取算法、利用椭圆几何对称性算法等).比较这些算法并选取几种分别应用于真实图像,并列出检测结果.最后,提出存在的问题和进一步研究的切入点.

Comparative Study on Ellipses Detection

Detecting ellipses from a digital image is very important in shape recognition and computer vision.Image edge detection is the first step to obtain image feature(including ellipses detection).In this paper,several approaches to edge detection such as Sobel operator,Canny operator,Zernike moment operator and so on,are briefly described firstly.We analyze the similarities/differences and advantages/disadvantages among them.Then,we emphatically describe several methods of ellipses detection based on Hough transform (HT) and its extensions.They are the method of using geometric properties,and the method based on techniques from Projective Geometry,and the detector using geometric symmetry,and so on.After interpreting and comparing the details on them,several representational approaches have been applied to real image.The results are shown after that.Finally,we address the problems and the issues for further research.