PVO based reversible data hiding technique for roughly textured images

Recently, several reversible data hiding (RDH) techniques based on pixel value ordering (PVO) have been proposed that precisely embed the secret data into cover images. As the neighboring pixels in smooth images are highly correlated, these methods perform better for smooth images but achieve comparatively low performance for roughly textured images. In many application domains like satellite imagery, the cover images are not always smooth. So, the performance of existing PVO based methods for roughly textured images needs to be improved. In this paper, we propose a novel RDH method based on PVO that is specially designed for improving the hiding performance in roughly textured images. In the proposed method, a segmentation scheme is used to cluster the pixels into different segments based on their intensity values. The segmentation ensures that the pixels in each segment are highly correlated to each other and each segment is divided into non-overlapping blocks of size $$ 2 \times 2 $$ where a block can hide at most two data bits in the smallest and the largest valued pixel. The size of the block is further extended by $$ 2 \times 1 $$ pixels if the complexity level of the block is ‘0’. The proposed method results in an increase in the hiding capacity as well as the visual quality of the stego images as the correlation of each block is increased which in turn limits the number of shifted pixels. The experimental results also prove the superiority of the proposed method against the existing PVO based RDH methods.