Nonrigid Correspondence and Classification of Curves Based on More Desirable Properties

We present a few more desirable properties to find correspondence and dissimilarity of two plane curves in nonrigid sense. A crossed scheme is used to define dissimilarity metric, which ensures an actual bi-morphism between two curves to be aligned. The optimal correspondence is found by a modified dynamic-programming method. From the optimal correspondence between two curves, we compute their dissimilarity by the overall difference of curvature, local dissimilarity, and local scale of stretching of their corresponding segments. Based on dissimilarity, we can do pattern recognition in nonrigid sense among curves which is very important in many areas such as recognition of hand written letters and cardiac curves where rigid transformations and scaling do not work well. Finally, the effect of correspondence and the classification is shown in application to cardiac curves.