Reversible data hiding algorithm based on adaptive prediction and code division multiplexing

A good predictor is crucial for reversible data hiding (RDH). At the same time, the RDH based on code division multiplexing (CDM) has been extensively researched and developed due to its outstanding performance. This paper proposes a high performance RDH algorithm based on adaptive predictor and code division multiplexing. The least square (LS) predictor is firstly employed to predict the target pixels according to the local consistency of pixels adaptively. The most correlated pixels are selected to enable the presented predictor to achieve high pixel prediction accuracy and efficiency, so that, the prediction errors are minimized and a more sparse error image is generated. Then, the minimized prediction errors are further collected to form base vectors for data embedding, and the CDM algorithm is utilized to embed secret data into the cover image. Due to the orthogonality of the spreading sequences, most elements of the selected vector are mutually canceled, therefore, high image quality can be maintained even at large data embedding capacity. Experimental results show that the proposed method is superior to Ma's CDM-based RDH scheme, and can provide competitive performances compared with other state-of-the-art RDH schemes.