椭圆特征的快速提取

Fast Extraction of Ellipses

基于椭圆和圆的对称性，对传统的霍夫变换检测曲线方法进行改进，提出分三步快速检测椭圆和圆的方法：（１）利用椭圆的几何对称性，快速确定可能的形心，生成相应子图，并滤除图象中的噪声；（２）在各子图中由边界点梯度向量和候选形心坐标，获得椭圆旋转角和长短轴比率；（３）根据椭圆参数方程确定椭圆的长短轴，并验证所检测到的椭圆．实验结果表明，该方法有效地降低了霍夫变换参数空间的大小和运算时间，提高了检测精度，实现了椭圆特征的快速提取．

Based on global geometric symmetry, the original Hough transform (HT) is improved, and a 3 stage approach to extract ellipses or circles is presented. The proposed method first locates candidates of ellipses or circles centers, then groups all the feature points into several subimages. And by using the gradient vectors of the feature point, we can obtain rotate angle, major diameter and minor diameter in each subimage. The method significantly reduces the time and space required to evaluate all possible parameters and improves the accuracy. The experimental results are given to show the correctness of the method.