Research On Image Steganography Algorithm Based On Lossless Compression And Low Pass Filtering

Existing image steganography algorithms mainly towards following challenges:(1)The performance is worse while using a mean filter to process the isolated noise and the edges of changed cost matrix.(2)When the secret message is embedded in the cover-image,the changed direction of the stego-image pixels is always random ?1,and the changed pixels are relatively scattered.It does not conform to the clustering rule in the design principle of steganography algorithm.Therefore,in order to design an image steganography algorithm with higher security,this paper improves two shortcomings of the existing image steganography algorithms,and proposes a new image steganography algorithm with higher security.The innovation of this paper is as follows:(1)Proposing a changed cost spreading method based on dynamic median filter.This method can replace the original mean filter to effectively deal with the isolated noise and texture edge in the changed cost matrix.(2)Proposing a new clustering method Clustering improved by LQ(Cluster-LQ)based on minimum filtering and perceptual hash algorithm.This method can effectively find the clustering region in changed pixels when the secret message is embedded,and make the regions with similar texture havethe same changed direction.(3)Based on the above research,proposing an image steganography algorithm Minimizing the Power of Optimal Detector improved by LQ(Mi POD-LQ),which is based on lossless compression and low pass filtering.This algorithm uses lossless compression method to compress secret message,uses the Mi POD's variance estimator of detector calculates the modified cost matrix of the carrier image,uses the dynamic median filter to smooth the changed cost matrix,then uses the clustering method Cluster-LQ to determine the changed direction of the changed pixel in the stego-image,and finally uses STC coding to embed the secret message.The experimental results show our algorithm Mi POD-LQ made the testing error rate of Spatial Rich Model(SRM)and max Spatial Rich Model(max SRMd2)be improved,which shows that the algorithm proposed in this paper has higher security than the current similar algorithm.