Abstract:H.265/HEVC-based video steganography has become a significant research direction in the field of information hiding. Typically, various syntax elements in the encoding and decoding process are exploited as carriers for video steganography, among which algorithms based on the coding unit (CU) block structure have emerged as a research focus. However, due to the limited number of CU partition modes, such algorithms often face the challenge of insufficient embedding capacity. Moreover, modifications to CU structures during the embedding process may lead to video quality degradation and increased susceptibility to steganalysis detection. To address these issues, this study introduces a novel video steganographic algorithm based on the improved deep texture-fidelity network (IDTFN) and multiple distortion function. Firstly, CUs of sizes 16×16 and 8×8 in I-frames are selected as embedding carriers, and a mapping scheme between CU partition modes and secret information is designed. Subsequently, a multiple distortion function is developed to guide the embedding process using syndrome-trellis codes (STC). Finally, to further optimize the quality of stego videos, IDTFN-based filtering is introduced for the first time in block-structure steganography. Experimental results demonstrate that, in comparison with the state-of-the-art CU block structure-based steganographic methods, the proposed algorithm achieves superior visual quality and enhanced resistance to steganalysis. Furthermore, it offers higher embedding capacity and more efficient bit utilization.