Efficient representation of size functions based on moments theory

Nowadays, there is a need to develop efficient and intuitive solutions such as hand gestures recognition for the Human-machine interaction. This paper presents a hand gestures recognition system based on salient geometric features extracted using size functions theory. We propose a new representation of the reduced size function based on Tchebichef moments providing more details and information for their graphs descriptions compared to existing representations. In addition, a methodical algorithm of fast Tchebichef moments computation for grey scale images is well adapted to the encoded graph of size function. Furthermore, a contour discretization based on a convexity approach is proposed for an optimal computation of the measuring functions, and new measuring functions for hand gestures classification and retrieval are proposed. The comparison with existing systems indicates that our method competes with the best ranked method for the dynamic case and surpasses the state of the art in static case; in addition it presents the advantage to be applied in both static and dynamic cases.