Cellular Neural Networks with second-order cells: Dynamics analysis and linear filtering

In this paper an alternative CNN model is proposed, in which the cell - the elementary processing unit of the array - is a second-order dynamic system. The dynamic behaviour is analyzed using the state equations. Concerning applications, some image linear filtering tasks are discussed and compared to the processing capabilities of the standard CNN model. As regards its pattern formation capabilities, the system eigenvalues are studied and we show how template and circuit parameters can be varied in order to obtain dispersion curves with a desired shape. As shown, this allows one to select the unstable modes and therefore to control pattern formation.