Image restoration using Lagrange programming neural networks

Summary form only given, as follows. An approach involving the use of Lagrange programming neural networks to restore degraded images is proposed. Three kinds of neurons take part in the restoration process: parameter neurons, variable neurons, and Lagrange neurons. The parameter neurons adapt to various input parameters, while the variable and Lagrange neurons construct a canonical computational circuit to fulfill fundamental solution-finding computation. Maximum entropy restoration is completed with ease by taking advantage of the features of robust stability and functional flexibility of the neural network. High-quality images were obtained in experiments.<<ETX>>