Statistical adaptive reversible steganographic technique using bicubic interpolation and difference expansion

Abstract

The reversible steganographic technique allows extraction of secret messages and restoration of original images without any distortion from the embedded image. In this work, a statistical adaptive reversible steganographic technique is proposed to improve difference expansion (DE)-based schemes, consisting of two parts. First, bicubic interpolation is adopted as the pixel prediction to obtain more embeddable pixels. Meanwhile, since differences are generated between the accurate predicted value and its original value, quality of difference is also considered. Second, a statistical adaptive reversible embedding algorithm is proposed to overcome the restriction of the embedding capacity under single-layer embedding. The relationship between the complexity of the neighboring pixels and the difference distribution for the image is generalized as the variance conditional in statistics. With the maximum modifiable degree of the predicted pixel, the proposed scheme provides a suitable embedding capacity for all embeddable pixels with less additional information. The experimental results demonstrate advantages of the proposed scheme and prove that it is able to provide high capacity with good visual quality for the embedded image.