Regularization-based continuous-time motion detection by single-layer cellular neural networks

Regularization theory is proposed for systematic design of linear- and nonlinear-connection-based cellular neural networks (CNN). In this paper, after stating the basics of regularization-based design of CNNs, such methodology is applied to the problem of continuous-time motion field estimation in moving images. A single-layer solution is thus obtained and simulated, paving the way to full two-dimensional focal-plane real-time motion detection circuit implementation.