A known-energy neural network approach for visual cryptography

This paper introduces the so-called known-energy system based on the Q'tron neural network (NN) model, and applies it to visual cryptography. With a known-energy system, the NN intrinsically performs a goal-directed search, meaning that the NN will settle down only when its state fulfils the dedicated goal. The noise injection mechanism that makes the NN to work in such a manner is discussed. The NN built for visual cryptography in the paper is modeled as a known-energy system. The approach is completely different from the traditional ones, and the so-built NN can be used to cope with complex encrypting structures of visual cryptography. Experiments show that its result is good.