Researches On Algorithm Of H.264 Video Steganography And Its Embedded Platform Implementation

In recent years, with the development of computer information and network, video steganography which is a new technology of information security develops quickly. H.264 is latest video coding standard and its compressive efficiency is the highest of the current video coding standards. But video steganographic algorithms about this standard are few. The good performances of H.264 are at the expense of computational complexity. So a general purpose MPU cannot satisfy the requirement of the H.264 real-time application. DSP chips are becoming ubiquitous with faster speed and lower resource consumption. Due to this ability, it has been used in this field.This thesis concentrates on the researches of H.264 video steganography and its implementation on TMS320DM642 for CIF resolution video. The main work is summarized as follows.First, a novel steganographic algorithm based on the phase of the motion vector is proposed. The algorithm utilizes the phase of single motion vector to embed the secret data, and use the matrix encoding technique to improve the embedding efficiency. Simulation experimental results demonstrate that the proposed approach can enlarge the steganographic capacity and reduce the modification ratio.Second, a real time steganographic algorithm by exploiting the IPCM (Intra-frame Pulse Code Modulation) macroblocks in P frames is proposed. This algorithm combines with the IPCM mode in which sample values are encoded directly withouting the need to predict and quantify. We use dynamic threshold to select IPCM macroblocks and use motion vector prediction value as the embedding intensity control information. Simulation experimental results demonstrate that the algorithm can get well embedding efficiency when the QP is smaller. Both subjective quality and objective quality are upgrade. This algorithm supports for blind extraction and meets real-time system requirements.Third, the thesis transplants the H.264 video steganographic algorithm into TMS320DM642 and optimizes it greatly. The performance of encode is great improved by using an excellent data transfer method based on EDMA (Enhanced Direct Memory Access) and PING-PONG technology.Fourth, completely portable software using for decoding and extraction is developed with SDL (Simple DirectMedia Layer).