Constrained multivariate linear regression adaptive image reversible information hiding error prediction optimization method

Abstract

The invention provides a constrained multivariate linear regression adaptive image reversible information hiding error prediction optimization method. The method comprises: selecting a prediction sample and a training sample, replacing a maximum value and a minimum value in the training sample with average values of training sample vector elements, establishing a multiple linear regression mappingrelation function matrix between training sample pixels and prediction sample pixels in a local area, and solving a linear regression function to obtain a training coefficient; and performing replacement optimization on the prediction sample by using the multivariate linear regression coefficient obtained by using the least square method and the prediction sample prediction target pixel value, and performing prediction of the target pixel value by taking the optimized prediction sample pixel as an independent variable sample.

Description

technical field [0001] The present disclosure belongs to the field of image information processing, and in particular relates to a constrained multiple linear regression adaptive image reversible information hiding error prediction optimization method. Background technique [0002] The information hiding technology realizes the transmission of secret information by hiding the information to be protected in the public carrier and transmitting the signal with the public carrier. Most of the information hiding algorithms will cause irreversible distortion of the original carrier, but in sensitive image processing fields such as military, medical, remote sensing, etc., it is unacceptable that the embedded information causes the distortion of the original image, thus resulting in the reversible information hiding algorithm. The image embedded with information based on the reversible information hiding algorithm can not only completely extract the secret information embedded in th...

AI Technical Summary

Problems solved by technology

[0005] However, as far as the inventor knows, traditional error prediction algorithms are based on calculating the similarity between the target pixel and its neighboring pixels, and the prediction of the target pixel value is realized through different arithmetic combinations of neighboring pixels. The consistency relationship between neighboring pixels in the image is still underutilized

Images