Detection of an ellipse by use of a recursive least‐squares estimator
Tadashi Nagata, Hirotoshi Tamura, Kohichi Ishibashi · Journal of Robotic Systems · 1985
Abstract A method for detecting an ellipse in an image by use of a recursive least‐squares estimator is described. This method is based on classifying the behavior of parameters in a quadratic equation estimated by the estimator. The segmentation of an image is executed using a statistical measure called the adjacent separability factor. Noise elimination is performed by the expansion‐contraction method and a contour for each region is extracted. Points on the contour are supplied to a recursive least‐squares estimator in succession and the parameters in a quadratic equation are estimated recursively. It is possible to decide whether the contour is an ellipse by analyzing the behavior of an estimated parameter. A main feature of this method is the generality and the simplicity of the procedure.