On aligning shapes

In this paper, we exploit some global shape properties in an efficient shape matching and retrieval method, which is robust with respect to rigid transformations (translation, rotation, scaling) and some deformations. The retrieval of similar shapes is obtained by normalizing the query shape and the database shapes using the area, the center of mass and the minimum bounding box. Then, an alignment is achieved by overlapping the shape areas. The retrieval of similar shapes is based on a similarity measure defined by the symmetric difference. The experimental results validate the proposed approach.