Image restoration using the RBF network with variable regularization parameters

摘要

This paper proposes a noise reduction filter that can suppress noise components without destroying important image information. The RBF network is then used to recover a high-quality image from the degraded version, and the regularization parameter is adjusted according to local image characteristics. A fast method of computing the RBF network with variable regularization parameters is developed by means of the properties of Kronecker product.