Learning Assisted Macro Block Selection for IPCM Mode Steganography in H.264 Videos

Among multiple modes for data embedding in H.264 video, IPCM mode has many benefits like higher embedding capacity and easy integration in H.264 encoding process. In most IPCM based data embedding solutions, a macro block is encoded into IPCM block only based on the amount of data to be embedded without considering the resulting quality of distortion after data embedding in the corresponding IPCM block. In this work, a learning assisted macroblock selection algorithm is proposed to select the macroblocks for IPCM encoding with the goal to minimize the quality distortion due to data embedding. In addition to selection of macroblock, the proposed solution also provides the number of least significant bits to be affected in the IPCM block encoding for the selected macroblock. Through performance testing, the proposed solution was found to provide 2.95% higher embedding capacity and 9.2% lower distortion compared to existing works.