Shape Edit Distance on Contour based Shapes

We present a new distance for the contour-based shapes recognition which we called shape edit distance (SED). Unlike the traditional distance, SED takes into account the spatial deformation of shapes. To compute the shape edit distance, we convert shapes to string sequences so that string matching techniques can be used. We pose the problem of string matching as a maximum a posteriori probability (MAP) alignment of the string sequences for shapes and compute the shape edit distance by finding the sequence of string edit operations which minimizes the cost of the path traversing the edit lattice. Experiments demonstrate the utility of the shape edit distance on a number of image retrieval and clustering problems