Adaptive HEVC video steganography based on distortion compensation optimization

Abstract

With the development of digital video communication, the steganography based on H.265/HEVC has attracted more and more attention. Because of higher security and better compatibility, the traditional video steganography often used quantized discrete cosine transform (QDCT) or quantized discrete sine transform (QDST) coefficients to embed secret messages. However, the QDCT/QDST-based steganographic methods will cause distortion drift seriously. Although the distortion compensation optimization method can eliminate the intra-frame distortion drift, it ignored the influence of inter-frame distortion. This limitation prompts authors to propose an adaptive HEVC steganography, which eliminates the intra-frame distortion drift while reducing inter-frame distortion. First, the distortion of intra-block and inter-block is analyzed with the prediction and transform process. Then, the corresponding embedding influence is defined, which includes intra-block distortion, inter-frame distortion accumulation, the block texture complexity and the cost of coefficient groups. Finally, the proposed adaptive cost function is designed based on the above definition and the syndrome trellis codes (STC) is used for embedding. The experimental results show that the proposed method has better performance compared with the state-of-the-art coefficient-based steganography in visual quality, bit rate increment and anti-steganalysis.