Radial basis function network for elimination of noise from image

A radial basis function network (RBFN) with under-relaxation iterative learning algorithm is proposed for the elimination of noise from image. This function is infinitely differential and locally supported. For a N×N image, the computational complexity of the learning algorithm is O(N²), which can be implemented using less storage and computation than the orthogonal least squares learning algorithm. Moreover, it may be used as an adaptive filter in computer vision and real-time signal/image processing. The test results show that the RBFN is suited for either Gaussian noise reduction or noise with uniform distribution.