Robust symbolic representation for shape recognition and retrieval

A new method for shape recognition and retrieval is proposed here. The suggested algorithm is based on several steps. The algorithm analyzes the contour of pairs of shapes. Their contours are recovered and represented by a pair of NN points obtained by linear interpolation. Given two points pipi and qjqj from the two shapes the cost of their matching is evaluated by using the shape context and by using dynamic programming the best matching between the point sets is obtained. Dynamic programming not only recovers the best matching, but also identifies occlusions, i.e. points in the two shapes which cannot be properly matched. Given the correspondence between the two point sets, the two contours are aligned using Procrustes analysis. After alignment, each contour is transformed into a string of symbols and a modified version of edit distance is used to compute the similarity between strings of symbols. Finally, recognition and retrieval are obtained by a simple nearest-neighbor procedure. The algorithm has been tested on a large set of shape databases (Kimia, MPEG-7, natural silhouette database, gesture database, marine database, swedish leaf database, diatom database, ETH-80 3D object database) providing performances for both in recognition and in retrieval superior to most of previously proposed approaches.