Direct least square fitting of ellipses

This work presents a new efficient method for fitting ellipses to scattered data. Previous algorithms either fitted general conics or were computation ally expensive. By minimizing the algebraic distance subject to the constra int 4ac - b(2) = 1, the new method incorporates the ellipticity constraint into the normalization factor. The proposed method combines several advanta ges: It is ellipse-specific, so that even bad data will always return an el lipse. It can be solved naturally by a generalized eigensystem. It is extre mely robust, efficient, and easy to implement.