Ellipse fitting with uncertainty and fuzzy decision stage for detection. Application in videomicroscopy

Abstract

The conic fitting from image points is a very old topic in pattern recognition. We propose here some new ideas of handling the difficult situations where the noised data span a small section of the conic. This new fitting process takes into account explicitly the maximum and minimum total arc length of the conic curves to constraint the conic search space. Our algorithm compared with some procedures of reference gives improved results. A confidence envelope is then estimated to direct the search for continuations of the ellipse. Considering the data organization resolved, we propose then a complete extraction scheme based on a fuzzy set representation of the fitting.