Capacity improvement of reversible data hiding scheme through better prediction and double cycle embedding process

Abstract

For better protection of surreptitious information from the intruders, steganography is now a widely used technique for hiding information through bits embedding inside a picture, speech and other multimedia contents. Among many different steganography schemes, prediction errors based reversible data hiding methods are more effective for security reasons. Nevertheless such techniques presented so far can embed limited number of message bits in a cover media and thus limit the length of covert message to be sent through embedding. In this paper, the authors presented a novel embedding process, called double cycle embedding scheme, where embedding capacity is increased significantly by modifying the scheme of estimation of pixel values, the block-variance calculation technique, selecting two highest most peaks in the histogram regardless their positions and embedding twice in the same cover image. The scheme also modified the pixel-prediction technique for the better predictions and for generalization of block-sizes in the cover image to increase the robustness of the policy. The experimental results demonstrate that the proposed scheme dictates other competing schemes with an embedding gain of 40%~700% depending on the cover image. The experimental results also demonstrate that the scheme exhibits stronger resistance against several statistical attacks, e.g. Benford generalized laws, SPAM features.