Research On Data Embedding Methods Based On Grayscale Image

With the rapid development of the Internet, informationalization has been a tendency, and human beings has started a fire-new age - Information and Multimedia Age. The way of getting information is not merely localized in conventional sheet, negative, tape and other formats. Multimedia and Internet have become more and more popular and efficient because of the easy-to-store and convenient-to-transmit of digital information. Although the digital technology brings us the convenience and business opportunity, some potential problems, such as the protection of privacy, copyright or the security of data transmission, have become more and more important. To solve the above problems, a new technology named data embedding emerges as the times require and has got sufficient development. Data embedding is that embeds secret information into media, such as texts, images, audios and videos, to avoid attracting the attention of interceptors, and enhance the security of data transmission. According to the utilization, this dissertation proposes the reversible and the non-reversible data embedding methods, and proposes an information Embedding method in compressed domain.(1) To achieve high quality stego-images, and get enough embedding capacity, a novel adjustable data hiding method was propose. The proposed method, based on exploiting modification direction (EMD), is called the "multi-group exploiting modification direction" method. In the multi-group method, combine several pixel-groups to embed secret data according to the constructed switch map to simply avoid the conversion redundancy of the EMD method and the spatial redundancy of the section-wise strategy, as well as to efficiently increase the probability of achieving larger embedding capacity. According to the experimental results and mathematical justification, the multi-group method can achieve higher embedding payload and better visual quality of the image than the EMD method and section-wise strategy.(2) According to the high correlation regulation between adjacent pixels, there were two new reversible data hiding methods proposed, one based on the high correlation between adjacent pixels to partition the image into blocks, and one with the direction order to classify the pixels. For different types of pixels, separately apply the interpolation prediction and the difference between adjacent pixels to embed the secret data into the corresponding pixels. In both methods, the mechanisms of processing the different types of pixels were different, and the orders of using interpolation prediction and difference between pixles were different, so the embedding performances were different. Yet both methods can provide the larger payload and the better image quality than some existing schemes.(3) Based on VQ, a reversible data embedding method was proposed. First, the original image was compressed into index table by traditional VQ. The index table was encoded by Huffman coding according to the weight of each index. Then, embed secret data into the encoded index table. Besides, according to a data structure for transmission, propose two schemes to deal with extra information. Experimental results show that the proposed method has superiority over Yang et al.’s method in the embedding capacity and the compression rate. Moreover, the embedding capacity or the compression rate can be adjusted freely according to actual application conditions.