Ellipse detection based on improved-GEVD technique
Yang Zhong-gen, Gui-Xiang Jiang, Lei Ren · 2004
The standard generalized eigen value decomposition (GEVD) is a popular ellipse detection technique whose statistical analysis is given to prove its disadvantages of very big estimation bias and MSE. It is also proved that the effective measurement to improve the performance of ellipse detection is whitening the data noise and regularizing data observation. This theoretic analysis has strongly supported the Hartley's regularization method. Then, an improved-GEVD algorithm has been developed. The theoretical analysis and computer simulation experiments have demonstrated that the proposed technique has the advantages that it is intrinsically able to whiten the data noise and to regularize the data observation so as to output a non-biased estimation of ellipse parameter with very small MSE. Furthermore, the computation complex is largely simplified.