Abstract
Neural-network approach to the restoration of blurred and/or noisy radiographic images is presented. Basic study was performed using a flower image, which was blurred by a Gaussian filter. A multi-layer, feed forward neural network with a back-propagation algorithm was designed to restore the degraded images. The effects of the network parameters, such as the number of neurons in each layer, on the restoration capability were experimentally investigated. The digital image restoration system based on the neural network was successful to improve the image quality with optimum network parameters. The structure consisting of 5×5 input neurons,10 hidden neurons, and one output neuron was most effective in our study. The network applied to the chest radiographs also showed a good performance. It was found that additional noise decreases the restoring capability. The preliminary results demonstrate the potential of an artificial neural network to restore the degraded radiographs.
