A New Descriptor for Shape Recognition and Retrieval

A new descriptor for shape recognition and retrieval is presented in this paper. The contour points of the shape are evenly sampled. With adjacent points linked by a line, the shape is represented by a complex polygon. The centroid of the polygon can be calculated with very low time cost. For every contour point, a two dimensional histogram is constructed to describe the spatial distribution of contour points with respect to the centroid and the point. Based on these 2D histograms, a three dimensional histogram i.e. the proposed descriptor is extracted. The descriptor is proved to be invariant to the similar transformation. The approach performs well on the experiments of shape recognition. And the experimental results of image retrieval show that it outperforms other methods.

[1]  Thomas Bernier,et al.  A new method for representing and matching shapes of natural objects , 2003, Pattern Recognit..

[2]  Su Yang.  Symbol Recognition via Statistical Integration of Pixel-Level Constraint Histograms: A New Descriptor , 2005, IEEE Trans. Pattern Anal. Mach. Intell..

[3]  Daphna Weinshall,et al.  Flexible Syntactic Matching of Curves and Its Application to Automatic Hierarchical Classification of Silhouettes , 1999, IEEE Trans. Pattern Anal. Mach. Intell..

[4]  Guojun Lu,et al.  Review of shape representation and description techniques , 2004, Pattern Recognit..

[5]  Philip N. Klein,et al.  Recognition of shapes by editing their shock graphs , 2004, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[6]  Beng Chin Ooi,et al.  Retrieving Similar Shapes Effectively and Efficiently , 2003, Multimedia Tools and Applications.

[7]  Benjamin B. Kimia,et al.  Symmetry-Based Indexing of Image Databases , 1998, J. Vis. Commun. Image Represent..

[8]  Jitendra Malik,et al.  Shape matching and object recognition using shape contexts , 2010, 2010 3rd International Conference on Computer Science and Information Technology.

[9]  Ulrich Eckhardt,et al.  Shape descriptors for non-rigid shapes with a single closed contour , 2000, Proceedings IEEE Conference on Computer Vision and Pattern Recognition. CVPR 2000 (Cat. No.PR00662).