Abstrakt

Hand Gesture Recognition using MACD

Geethu Prakasan, Bimal Bose B.S

Shape disintegration is a principal issue for part-based shape representation. We propose the minimum approximate decomposition (MACD) to deteriorate arbitrary shapes into least number of "approximate convex” parts. The minimum approximate decomposition is detailed as a discrete optimization problem by minimizing the amount of nonintersecting cuts. Two discernment guidelines are forced as constraints into the objective function to enhance the visual naturalness of the decomposition. With the level of close convexity a userspecified parameter, the proposed decomposition is robust to local distortions and shape deformation. The optimization could be effectively explained by means of binary integer linear programming. Both theoretical analysis and experiment results indicate that our approach outperforms the state-of-the-art results without presenting redundant parts and in this way leads to strong shape representation.