2D Shape Recognition Using Information Theoretic Kernels

In this paper, a novel approach for contour-based 2D shape recognition is proposed, using a recently introduced class of information theoretic kernels. This kind of kernels, based on a non-extensive generalization of the classical Shannon information theory, are defined on probability measures. In the proposed approach, chain code representations are first extracted from the contours; then n-gram statistics are computed and used as input to the information theoretic kernels. We tested different versions of such kernels, using support vector machine and nearest neighbor classifiers. An experimental evaluation on the chicken pieces dataset shows that the proposed approach outperforms the current state-of-the-art methods.

• Type: Conference
• Year: 2010
• Where: ICPR
• Authors: Manuele Bicego, André Filipe Torres Martins, Vittorio Murino, Pedro M. Q. Aguiar, Mario A. T. Figueiredo
• Keywords: Computer Vision, Contour-based 2D Shape, ICPR 2010, Information Theoretic Kernels, Nearest Neighbor Classifiers