Two-level embedding reversible data hiding algorithm based on two-dimensional biased distance interpolation
A double-layer embedding reversible information hiding scheme based on prediction error extension is proposed to address the issue of the inability to balance the embedding rate and distortion of encrypted images in current reversible information hiding technologies.An improved omnidirectional gradient parameter combined with a two-dimensional biased distance interpolation algorithm is used to divide the image pixels into two groups in a chessboard manner by the first layer of embedding.One group of pixels is used to predict the other group of pixels,and then secret information is embedded according to the prediction error expansion method.The second layer of embedding takes the remaining set of pixels as the target pixels,and then performs prediction and embedding to further improve the embedding rate.Experimental analysis from multiple aspects such as algorithm reversibility,prediction accuracy,embedding capacity,and carrier image distortion is conducted in this paper.The results show that the peak signal-to-noise ratio of the restored image after double-layer embedding is ∞,and the mean square error of different types of hidden image pixels and restored image pixels is 0,indicating that the algorithm can achieve completely reversible hiding.Testing different types of carrier images showed that 79%of the tested images had an average prediction error of less than 4,indicating high prediction accuracy of the algorithm.The comparison experiment of embedding rate shows that compared with other algorithms,this algorithm has the highest embedding rate,and the average improvement of double-layer embedding rate is more than 30%compared to single-layer embedding,fully utilizing pixel redundancy.When the embedding rate is 0.7 bit/pixel,the average peak signal-to-noise ratio of the encrypted image and the original image in this algorithm is 39.854 dB,which is 3～5 dB higher than similar algorithms,indicating lower distortion of the encrypted image.This article also conducts analysis and comparative experiments on the correlation between threshold and embedding rate,as well as structural similarity and embedding rate,to deeply explore the conditions for obtaining high embedding rate.This algorithm can effectively solve the contradiction between imbalanced embedding rate and distortion of encrypted images,and has better overall performance.
reversible data hidingdifferential prediction expansion2-D biased distance differenceomni-directional gradienttwo-level embedding
国家科技期刊平台
NSTL
万方数据
