彩色图像中椭圆的快速检测算法  

Fast Ellipse Detection Algorithm In Color Images

作　　者：华继钊[1] 郭志波[1] 陈才扣[1] 杨静宇[1] 

机构地区：[1]南京理工大学计算机系

出　　处：《工程图学学报》2007年第5期82-87,共6页Journal of Engineering Graphics

基　　金：江苏省自然科学基金资助项目(BK2002001);图像处理与图像通信实验室开放基金资助项目(IPICL0304)

摘　　要：椭圆检测是图像处理中常用的技术,由于自然界很多物体都可以用椭圆进行拟合,所以也成为图像分割和目标提取的关键技术。该文分析了椭圆的解析结构,利用椭圆长短轴之间的几何关系,从边缘点集合中选取全局性的参数,快速生成椭圆。算法针对彩色图像,引入了彩色边缘增强算法和基于图搜索的边缘追踪算法,生成封闭的边缘轮廓曲线,对每段曲线进行椭圆拟合,通过拟合评估,形成拟合判决,确定图像中的椭圆区域。算法计算量小,检测速度快,对多椭圆和椭圆弧都能准确检测。Ellipse detection is a common technique in image processing. Since many natural objects are similar to ellipse, so ellipse detection becomes one of the key techniques in image segment and object extraction. This paper analyses analytic architecture of ellipse, based on the geometrical relation between long axis and short axis of an ellipse, in the help of the overall parameters selected from the set of edge pixels, proposed a algorithm to generate an ellipse quickly. By using edge enhancing algorithm and edge tracing algorithm based on graph searching in color images, a curve composed of edge pixels is generated. Then every curve is fitted by an ideal ellipse. If the fitness passes the evolution, the curve is regarded as an ellipse and the region closed in the curve is an ellipse region. The algorithm proposed in this paper needs less computation, provides high detection speed, and can be used to detect full ellipse and arc of ellipse accurately.

关 键 词：计算机应用 椭圆检测 图像分割 目标提取 拟合度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]