OMDA: An open GOP-oriented multi-level distortion analysis-based adaptive video steganography method in the PU partition mode domain

The high efficiency video coding (HEVC) standard features a flexible block partitioning structure and rich coding syntax elements, providing a variety of cover element choices for compressed-domain video steganography. However, existing video steganography methods based on the HEVC block partitioning structure still have two shortcomings. On the one hand, block partitioning-based video steganography methods are prone to disrupting the original coding partitioning structure, easily leading to a decrease in compression efficiency. On the other hand, some methods mainly focus on local distortion within blocks, neglecting inter-block correlation and inter-frame distortion propagation, resulting in a decrease in visual quality and compression efficiency. To address these issues, we propose an open GOP-oriented multi-level distortion analysis-based adaptive video steganography method (OMDA) in the intra-frame prediction unit (PU) partition mode domain. OMDA selects the intra-frame PU partitioning mode with the smallest CU size as the cover elements while maintaining the coding unit (CU) partitioning structure in I-frames, and analyzes the distortion impact of video steganography operations at three levels: inner-block, inter-block, and inter-frame. Furthermore, OMDA constructs a distortion function based on the analysis results and finally combines Syndrome-Trellis Codes (STC) to achieve minimum distortion embedding, thereby reducing the impact of video steganography operations on the visual quality, compression efficiency, and security of stego videos. Experimental results show that our proposed OMDA has good overall performance in terms of visual quality, compression efficiency, and security.