Reversible data hiding in encrypted binary images by pixel prediction

Abstract

In this paper, reversible data hiding in encrypted binary images (RDHEBI) is proposed. Considering that there is not much redundancy in binary images and arbitrarily flipping pixels usually causes noticeable distortions, the original image is divided into non-overlapping blocks, including uniform blocks (pixels are all-white or all-black) and non-uniform blocks (both white pixels and black pixels exist). A type image is generated to identify these two kinds of blocks. After image rearranging and type image self-embedding, the preprocessed image is encrypted via a standard stream cipher. In the data embedding phase, the secret data are embedded into both uniform blocks and non-uniform blocks. Finally, by means of pixel prediction in “T” patterns, the receiver can obtain the secret data with the data embedding key, or the approximate original image with the encryption key, or the secret data and the original image with both keys. Moreover, the secret data can be extracted in both encrypted domain and decrypted domain. Experimental results show that there are no distortions with a low embedding rate or very few distortions with a high embedding rate. In addition, an extended lossless method is given to achieve complete reversibility independent of embedding rate.