Reversible data hiding in encrypted image based on multiple linear regressions and adaptive adjustment of the prediction data

Abstract

As an effective mean, reversible data hiding in encrypted images (RDH-EI) can be used for privacy protection by means of encryption and data hiding technology, and it can also be used to hide data in the encrypted image without any information of image content. In this paper, a large capacity RDH-EI algorithm based on multiple linear regressions (MLR) and multi-bit prediction (MBP) is proposed. In the scheme, the pixels of the original image are divided into two classes; the pixels in the first class is predicted by its afore pixel. The pixel in the second class is predicted by MLR through its four neighboring pixels, and then the label map of pixels (LMP) is adaptively generated based on the prediction error and smoothness of image. In this way, large capacity space for data hiding can be reserved before image encryption, and secret data can be embedded into the encrypted image by multi-bit substitution. Experiments and analyses on some standard test images and three image datasets show that the proposed scheme achieves the higher payload than those obtained with current state of the art methods, and the adjustment techniques is reasonable and effective.