A comparative study of three shape normalization algorithms

Shape is one of the fundamental visual features in pattern recognition and target tracking. And shape normalization is a very important pre-processing step in image understanding. In general, there are four basic forms of planar shape distortions caused by changes in viewer´s location: translation, rotation, scaling and skewing. A good shape descriptor should be invariant to these distortions. In this paper, we study and compare three shape normalization algorithms: J.G. Leu´s shape compacting algorithm and its two modified versions: Wang´s image ellipse algorithm and Liang J.J.´s principal axis algorithm. Experiments on a set of images show that all of these algorithms have some drawbacks and we give some advices for modification.