Inverse halftoning using a multilayer perceptron neural network

Digital halftoning is an active research theme, which can be applied in many fields of image processing. There are several methods with different characteristics. In digital halftoning, we perform the gray-scale to bi-level conversion using software or hardware and the inverse halftoning is a reconstruction technique of a gray-scale image from its halftone version. This paper proposes a new method for obtaining a gray-scale image from its halftone version. This method uses a Multilayer Perceptron neural network (MLP) trained by a Backpropagation (BP). A high quality of the gray-scale image obtained by the inverse halftoning is required in many applications. The proposed method offers a high quality of reconstructed gray-scale image, comparing with the previously proposed methods. The experimental results demonstrate the effectiveness of the proposed inverse halftoning algorithm.