Research On Optimization Algorithm For Large-capacity Reversible Information Hiding

Reversible information hiding is a kind of information hiding with special properties.The difference is that it will not cause permanent distortion to the host image,and can perfectly reconstruct the host image after extracting the secret information.Due to this feature,the reversible information hiding can be well applied in some special application scenarios,such as military,medical,satellite,law,treasure protection and so on.Based on the information theory framework of reversible information hiding,this thesis focuses on the optimization algorithm of large capacity reversible information hiding.Two different optimization algorithms are proposed,the contributions are as follows,Firstly,a reversible information hiding optimization algorithm based on the structural square error constraint is proposed.The structural square error is a kind of image quality evaluation index,which inherits the excellent mathematical characteristics of the square error and can extract the distortion of the image structure information.Compared to the traditional image quality evaluation indexes,such as PSNR and SSIM,it can reflect the human visual characteristics more.Thus,a reversible information hiding rate-distortion model under the structural squared error constraint is first established.Then,based on the quasi-Newton idea,the path following algorithm is improved,and generates a BFGS-PF algorithm.The experimental results show that the rate-distortion curve of the BFGS-PF and the path following algorithm is very close,but the solution speed has been greatly improved.Moreover,the reversible information hiding algorithm based on BFGS-PF can be well applied to the application scenario of large embedding rate.Secondly,a reversible information hiding optimization algorithm based on the linear minimum entropy prediction criterion is designed.The core idea of the minimum entropy prediction criterion is to predict the value of pixels in the perspective of information entropy,so as to ensure the high precision of the prediction value.First,we use the minimum entropy prediction to get a more steep prediction error histogram,and then use the improved recursive coding method to modify the prediction error histogram to achieve the best embedding of secret information.The experimental results show that the new algorithm has been effectively improved in two aspects of prediction accuracy and embedding performance.The above two algorithms not only improve the rate-distortion theoretical framework of reversible information hiding,but also achieve some success in enhancing pixel prediction accuracy and embedding capacity.