Neural network functional models and algorithms for information conversion in order to create digital watermarks
А. А. Сирота, M. A. Dryuchenko, E.Yu. Mitrofanova - · Radioelectronics and Communications Systems · 2015
There are considered functional neural network models and algorithms of information conversion that providing steganographic encoding of messages in the form of digital watermarks (DWM) into arbitrary objects—containers (digital images) and their subsequent decoding with minimal container distortion. The approach is based on theoretical justification of creating hetero- and autoassociative contraction mappings of the container fragments using direct propagation neural networks. The dependencies of the DWM quality indicators describing the container distortion level, as well as the probability of error at the DWM binary sequence decoding were obtained for model images in the form of random fields, as well as for real images.