Inverse Halftoning using Neural Networks based methods

Recently inverse halftoning techniques are applied in many image processing applications, in which an efficient inverse halftoning method, that provides high quality gray-scale image from any binary halftone image, is required. In this paper we propose two neural networks based inverse halftoning methods, which are Multilayer Perceptron (MLP)-based and Radial Basis Function (RBF)-based inverse halftoning methods. In both methods, the training stage is required using some halftone images and their corresponding gray-scale images, however once both neural networks are trained, the adapted connection weight values can be used to generate gray-scale images from any unknown halftone images in training stage. The proposed methods provide the higher quality gray-scale images compared with the previously reported methods, while keeping lower temporal and spatial complexities. Key-Words:Halftoning, Inverse Halftoning, Neural Networks, MLP, RBF, Gaussian, HVS