High Efficiency Reversible Data Hiding for Two-stage Vector Quantization Compressed Images

Data hiding is a booming technique aiming to embed secret data into digital media for content authentication, annotation or copyright protection. Reversible data hiding is a special type of data hiding technique that guarantees the host digital media can be losslessly recovered after the secret data are correctly extracted. In the past several years, some reversible data hiding schemes for Vector Quantization (VQ) compressed images have been proposed. Some of them embed data during the VQ encoding process, while others embed data during the VQ index streaming process. Compared with the traditional VQ, Multi-Stage Vector Quantization (MSVQ) is a more attractive data compression scheme with lower complexity and less encoding time due to the reduced codebook size. This paper proposes a novel reversible data hiding scheme based on two-stage VQ. The proposed framework consists of three phases, two-stage VQ codebook generation, path selection and data embedding, and cover image reconstruction and secret data extraction. Our main contribution lies in improving both the payload and transmission efficiency with a smaller codebook in comparison with traditional VQ-based reversible data hiding schemes. Experimental results validate the effectiveness of our scheme and demonstrate that it outperforms several previous methods in terms of payload, efficiency and stegoimage quality.