Prediction-Based Reversible Data Hiding for Medical Images with Genetic Algorithms

Reversible data hiding is a newly developed topic in watermarking researches. At the encoder, it relies on slightly modifying the characteristics of original images for embedding secret information. At the decoder, original image and secret information can be separated from marked image with slight amount of overhead. In this paper, we propose the scheme by predicting the difference between output and input images for making reversible data hiding possible. By carefully selecting prediction coefficients, which are optimized by genetic algorithm, the output image quality can be preserved, while the enhanced amount of embedding capacity can be observed. We apply the algorithm to medical images for protecting patients' cases from possible human errors incurred. With the training of genetic algorithm, simulation results with our algorithm have demonstrated the enhanced embedding capacity, while keeping the output image quality. Optimized prediction coefficients with genetic algorithm lead to better performances with our scheme.