Research On Steganographic Algorithms Based On Features Of Digital Image

Steganographic algorithm is mainly focusing on three indicators: perception,security and embedding capacity.It is import and difficult to design steganographic algorithm with high embedding capacity,low perceptual distortion and high security.These difficulties also make the development of steganography from least significant bit to multi-pixel embedding,and then it evolved into image-based steganography.In this article,three different types of steganographic algorithms have been proposed aim at different image features.They are mainly around the perception,security,embedding capacity and embedding time.At last,the research results are as follows:1.The steganographic method based on algebraic multigrid and diamond encoding is proposed.Firstly,the pixels of the image are divided into two parts with algebraic multigrid,and then the confidential information was embedded into the two sequences of the coarse and fine grid by diamond encoding respectively.Finally,the algorithm selects the k value of diamond encoding equal to 1 and 2,and propose three steganographic schemes.Under the condition of the same embedding rate,the experimental results compared with least significant bit replacement,least significant bit matching,original diamond encoding and adaptive edge detection steganographic algorithm,besides it uses image structure similarity and first-order Markov safety index to analysis,and it shows that this steganographic method can maintain imperceptibility and improve security.2.The steganographic method based on Canny edge detection is proposed.Because the edge of the image is changed with the Canny edge detection threshold,as well as the confidential information embed into edge pixel is not easily detect relative to the non-edge pixel.Thus it can use threshold iteration to maximize the edge pixel sequence of Canny edge detection,and then select the k value of diamond encoding is equal to 1 as the embedded algorithm.Eventually,at the same embedding rate,this proposed method has higher PSNR.In terms of security,firstly extract the feature of secret image and original image,and then use support vector machine to train and calculate minimum average decision error.The experimental results compared with least significant bit,original diamond encoding's k value is equal to 1 and adaptive edge detection steganographic algorithm,with embedding capacity growing it has higher security.3.Considering the embedding capacity and embedding time of steganographic algorithm,the edge detection steganographic algorithm is designed based on the performance.The embed principle is that replacing one or more bits of the pixels' least significant bit,and confidential information embedded into edge pixel is greater than non-edge.Finally,under the condition of guarantee the image perception,the maximum embedding capacity of edge detection steganographic algorithm is greater than 1.5bpp.At the same time,the embedding time is less than 1.55 seconds when the embedding capacity is 1.0bpp,and thus it reflects the practical value of the edge detection steganographic algorithm.