Research On Key Technology Of Video Information Hiding

As an effective method to realize copyright protection and covert communication, video information hiding has attracted much attention. Video information hiding based on visual perception characteristics has important research values and broad application prospects. The major problem of perception based information hiding is how to build the visual perception model and adopt it in video information hiding. This dissertation derives the information hiding and analysis method based on visual perception model from human visual system; in the light of wide application of compressed video, information hiding in compressed domain is also studied in this dissertation.The main work and innovation are as follows.1) The redundancy of Huffman codes in compressed domain is used to propose a blind information hiding algorithm based on VLC mapping. Through improving the way VLCs are paired and mapped, the projective relation between VLCs before and after mapping is obtained, which makes the blind extraction realized while preserving the file size.2) A novel image steganographic algorithm based on edge detection is presented. Sobel operator can be used to effectively discriminate the texture and smooth regions in image. It's modified to Gradient Template for Steganography in order to realize blind extraction. Secret information is embedded under control of capacity factor. Experimental results are given to show ability to hide secret data of large capacity without noticeable distortion.3) A novel high capacity video steganographic algorithm based on visual perception model is presented. This algorithm integrates multi-parameter spatial contrast sensitivity function, contrast masking based on classification of DCT blocks, and motion attention. Then secret information is embedded under control of distortion masking measure of each DCT coefficient. Experimental results are given to show ability to hide secret data of large capacity without noticeable distortion.4) A novel steganalysis method is developed against video steganographic algorithm based on visual perception model. Estimation frame of secret message is calculated using inter frame collusion based on block division. The estimated frame is compared with distortion measure to decide whether a secret message is hidden in the video. The classification of similarity of these two images is completed by sending the algebraic, texture and shape features into support vector machine. Experimental results show that the steganalysis method has good correct detection rate which is about 90%.