Irregular-closed graphic recognition based on contour measurement matrix

For higher efficiency of irregular-closed graphic recognition than the traditional algorithms, a self-defined contour measurement matrix is introduced, and based on which a fast recognition algorithms is proposed. The contour matrix optimally describes the geometric characteristic of the graphic outline, then a group of eccentric-rate parameters is defined based on the matrix, and the probability density function of the parameters is considered as the basis of classification. Combined with Bayes theory, a design method of classification with minimum-error probability (MEP) is proposed. The proposed algorithm is demonstrated by applications to be invariant to translation and scaling, is about 30 times more computationally efficient than the Fourier-based descriptors.