The stability of a class of delayed cellular neural networks (DCNN) with or without noise perturbation is studied. After presenting a simple and easily checkable condition for the global exponential stability of a deterministic system, we further investigate the case with noise perturbation. When DCNN is perturbed by external noise, the system is globally stable. An important fact is that, when the system is perturbed by internal noise, it is globally exponentially stable only if the total noise strength is within a certain bound. This is significant since the stochastic resonance phenomena have been found to exist in many nonlinear systems.

A new texture classification method based on singular value decomposition(SVD) and wavelet transform is presented. Wavelet transform is employed on texture images having been preprocessed with SVD. The elements of the signature vector of an image are the fractal dimensions and barycentric coordinates of the bit planes of the wavelet coefficients in both the 3-Level high frequency domains and the third low frequency domain. The one-nearest-neighbor classifier with standard L ଵ -norm distance is utilized to perform supervised texture classification. Compared with some other classification methods, the method is experimentally proved more efficient and less time-consuming.

No abstract